{
  "version": "https://jsonfeed.org/version/1.1",
  "title": "BCVR Dispatch",
  "home_page_url": "https://www.bcvr.io/blog/",
  "feed_url": "https://www.bcvr.io/blog/feed.json",
  "description": "Weekly notes on immersive production, XR, spatial sound and AI film from BCVR Immersive, London.",
  "icon": "https://www.bcvr.io/assets/roundel.png",
  "favicon": "https://www.bcvr.io/assets/roundel.png",
  "language": "en-GB",
  "authors": [
    {
      "name": "Scott Marshall",
      "url": "https://www.bcvr.io"
    }
  ],
  "items": [
    {
      "id": "https://www.bcvr.io/blog/the-house-pair/",
      "url": "https://www.bcvr.io/blog/the-house-pair/",
      "title": "The House Pair",
      "summary": "Camera glasses went on pre-order last week from $395. Cultural venues had already begun banning the visitor's own. The pair that survives is the one the venue hands over.",
      "content_html": "<p>On 12 November Google opened pre-orders for what it calls Intelligent Eyewear: Gemini glasses designed with Warby Parker, Gentle Monster and Samsung, <a href=\"https://blog.google/innovation-and-ai/technology/xr-ar/google-intelligent-eyewear-collections-pre-order/\" target=\"_blank\" rel=\"noopener\">starting at $395</a> and in shops from 10 December. The specification reads like a list of things that have been taken away. The frames weigh &quot;around 41 grams&quot;. The battery &quot;lasts up to nine hours&quot;. There is a 12-megapixel camera and, as <a href=\"https://9to5google.com/2026/11/12/google-intelligent-eyewear-glasses/\" target=\"_blank\" rel=\"noopener\">9to5Google reports</a>, a physical switch so that you can &quot;wear them as regular glasses only&quot;.</p>\n<p>That last phrase is the design goal of the whole category, and it is the difficulty at the door. Five weeks earlier, on 9 October, the cultural sector of Groningen published a <a href=\"https://www.kunstpuntgroningen.nl/blog/statement-cultuursector-groningen-geen-slimme-brillen-in-onze-gebouwen-en-tijdens-onze-evenementen/\" target=\"_blank\" rel=\"noopener\">joint statement</a> saying, in translation, that smart glasses with a camera and AI have no place in its buildings or at its events. <a href=\"https://nltimes.nl/2026/10/09/35-groningen-clubs-museums-events-jointly-ban-smart-glasses-privacy-infringements\" target=\"_blank\" rel=\"noopener\">NL Times put the number at 35</a>. The published list, which says it is not exhaustive, names 37, including the Groninger Museum, Forum Groningen, the Grand Theatre, Vera, Simplon, Eurosonic Noorderslag and Noorderzon. The signatories said they were writing a ban into their house rules.</p>\n<p>So in one autumn Google joined Meta in putting a camera into an ordinary-looking frame, and a city&#39;s museums, theatres and festivals declined to admit it. For anyone who builds immersive work in cultural buildings, the collision matters more than either event alone. For a decade the pitch has been that visitors would one day arrive already wearing the display.</p>\n<h2 id=\"a-rule-nobody-can-see-to-enforce\">A rule nobody can see to enforce</h2>\n<p>Groningen is one entry in a longer list. On 5 October the <a href=\"https://www.clickondetroit.com/tech/2026/10/05/norway-seeks-temporary-ban-on-smart-glasses-as-privacy-concerns-rise/\" target=\"_blank\" rel=\"noopener\">Associated Press reported</a> that Norway&#39;s government would propose a temporary ban in some places, one that &quot;could extend to public concerts and sports events, plus parks, beaches, museums and shopping centers&quot;. The government stressed that it was not proposing a total ban, and said a bill would go to parliament &quot;as soon as possible&quot;. In August the UK Cinema Association said many members were &quot;introducing policies to prohibit and/or restrict the wearing of camera-enabled smart glasses&quot;, and <a href=\"https://www.aol.co.uk/articles/piracy-fears-prompt-calls-ban-140632000.html\" target=\"_blank\" rel=\"noopener\">the Guardian reported</a> that ATG Theatres has said people wearing them during shows will be asked to remove them.</p>\n<p>Read these closely and they share a gap. Each describes a rule, or the intention of one, and none describes a check. Neither the Groningen statement nor the NL Times and OOG TV reports of it say how door staff would tell one black frame from another. The Guardian&#39;s report adds that the cinema association was not aware of any bans by members to date. VR.org&#39;s <a href=\"https://vr.org/articles/uk-venues-ban-meta-smart-glasses-wetherspoons-soho-house-2026\" target=\"_blank\" rel=\"noopener\">write-up of the UK venues</a> was blunt: &quot;Enforcement will be imperfect&quot;, because busy bar staff &quot;cannot reliably spot a pair&quot;.</p>\n<p>The manufacturers&#39; answer is the indicator light. Google says its glasses &quot;are designed to stop recording if the light is covered or disabled&quot;. Meta got there after a public backlash. In August <a href=\"https://www.engadget.com/2245776/meta-closing-loophole-that-allowed-people-to-record-with-smart-glasses-light-covered/\" target=\"_blank\" rel=\"noopener\">Engadget reported</a> Meta&#39;s Alex Himel announcing a fix so that &quot;the camera will now stop working if the light is covered during a recording&quot;, after wearers found they could start filming and then cover the LED. The same report describes &quot;a thriving market for cheap accessories that can bypass the recording light without detection&quot;. A light is a signal to the person standing in front of the wearer.</p>\n<blockquote class=\"pull\"><p>An indicator light tells front of house nothing about the third row of a dark room.</p>\n</blockquote><figure class=\"post-figure\"><img src=\"https://www.bcvr.io/blog/images/the-house-pair/figure-1.jpg\" alt=\"A visitor wearing dark-framed glasses stands before a large landscape painting in a red-walled gallery\" width=\"1200\" height=\"670\" loading=\"lazy\" decoding=\"async\"><figcaption>Nothing about the frame tells the gallery which kind of glasses these are.</figcaption></figure>\n<h2 id=\"the-exemption-is-the-interesting-part\">The exemption is the interesting part</h2>\n<p>The Groningen statement carries one sentence that does more work than the rest. Aids used for supportive or medical purposes, it says, with an &quot;of course&quot;, are not covered.</p>\n<p>Those aids are the same glasses. On 14 October, five days after Groningen, Meta, EssilorLuxottica and Milan&#39;s Pinacoteca di Brera <a href=\"https://about.fb.com/news/2026/10/meta-milans-pinacoteca-di-brera-and-essilorluxottica-expand-access-to-art-through-ai-glasses/\" target=\"_blank\" rel=\"noopener\">launched Sensing Art</a>. A blind or low-vision visitor stands in front of &quot;one of eight selected artworks on the itinerary&quot;, starts artwork recognition from the glasses and hears a description through the open-ear speakers. The tour was developed by the Associazione Nazionale Subvedenti using its DescriVedendo method. Meta&#39;s announcement has the glasses &quot;available Tuesday through Friday&quot;, by reservation.</p>\n<p>It never says outright that the museum hands the pair over, and <a href=\"https://mixed-news.com/en/meta-ai-glasses-pinacoteca-brera-sensing-art-eight-artworks/\" target=\"_blank\" rel=\"noopener\">MIXED noted</a> that Meta gives no end date, fee or visitor numbers. The shape is still legible: a fixed route, authored descriptions, a booked slot, a known device.</p>\n<p>The accessibility case deserves better than an &quot;of course&quot;. Google&#39;s own launch post promotes the Aira visual interpreting service on its glasses. Colin Hughes, a former BBC producer who campaigns on disabled people&#39;s access to technology, <a href=\"https://www.aestumanda.com/opinion/2026/07/smart-glasses-backlash-disabled-people/\" target=\"_blank\" rel=\"noopener\">wrote in July</a>: &quot;We should not have to choose between privacy and independence. We should insist on both.&quot; A door policy that depends on staff judging who looks as if they need their glasses will fail those visitors first.</p>\n<figure class=\"post-figure\"><img src=\"https://www.bcvr.io/blog/images/the-house-pair/figure-2.jpg\" alt=\"Typographic title card reading The House Pair, with a line drawing of glasses resting on a tray\" width=\"1200\" height=\"679\" loading=\"lazy\" decoding=\"async\"><figcaption>The pair a venue can vouch for is the one it issues.</figcaption></figure>\n<h2 id=\"what-the-house-pair-buys\">What the house pair buys</h2>\n<p>None of this is new in principle. The audio guide handset and the headset in a location-based VR attraction are house devices, cleaned, charged and counted back in. London&#39;s National Theatre has offered bookable <a href=\"https://press.epson.eu/en_GB/case-studies/national-theatre-uses-moverio-to-provide-captioning-for-hearing-impaired/\" target=\"_blank\" rel=\"noopener\">Epson smart caption glasses</a> to deaf and hard of hearing audiences since a 2018 trial. In August, Ningbo&#39;s Tianyi Pavilion was <a href=\"https://ningbo.chinadaily.com.cn/2026-08/06/c_1202944.htm\" target=\"_blank\" rel=\"noopener\">reported to be piloting Rokid glasses</a> that give &quot;automatic commentary based on visitors&#39; locations and surroundings&quot;.</p>\n<p>What the consumer launch changes is less than it looks. Google&#39;s footnote says Intelligent Eyewear is &quot;only available for use with Google accounts from the United States and South Korea&quot; and needs a compatible 18+ account. It is a personal device, bound to one adult&#39;s account and phone, so it is hard to lend, and it is not yet on sale in Britain at all. The visitor&#39;s own pair is being written out of the house rules, and the retail pair is not built to be handed across a desk. The gap between them is a fleet device a venue can manage.</p>\n<p>At BCVR we build hybrid physical and digital installations, and the device has always been on our kit list. Three things now argue for keeping it there. Generative tools have cut the cost of drafting and translating the authored layer. Brera&#39;s descriptions were still written by specialists, as they should be, so the money moves to the handover: charging, cleaning, fitting over a visitor&#39;s own lenses, the minute at the desk. An issued pair returns evidence a personal one never will, with consent and in aggregate: pairs out, routes finished, stops skipped. And a venue can say exactly what its own device records, which it cannot say of a stranger&#39;s.</p>\n<p>There is a rule to borrow as well. Royal Caribbean&#39;s policy, as logged in <a href=\"https://vr.org/smart-glasses-bans\" target=\"_blank\" rel=\"noopener\">VR.org&#39;s tracker</a>, reads: &quot;Smart glasses may be brought onboard; however, they may not be used in the following areas.&quot; It regulates use by zone. Staff still have to notice the glasses, but only where it matters.</p>\n<p>That tracker, updated on 15 November, still lists Norway&#39;s bill as yet to come. Google&#39;s glasses reach the shops on 10 December. Somewhere between those dates a front of house manager will be asked what the policy is, and the workable answer involves a tray of house pairs, offered alongside a visitor&#39;s own assistive glasses and never in place of them.</p>\n<p><em>Scott is the founder of <a href=\"https://www.bcvr.io\">BCVR</a>, a London studio producing immersive installations across XR, projection, spatial audio and generative AI.</em></p>\n<h2 id=\"sources\">Sources</h2>\n<ul>\n<li><a href=\"https://blog.google/innovation-and-ai/technology/xr-ar/google-intelligent-eyewear-collections-pre-order/\" target=\"_blank\" rel=\"noopener\">Google: See our intelligent eyewear lineup and pre-order your pair (12 Nov 2026)</a></li>\n<li><a href=\"https://9to5google.com/2026/11/12/google-intelligent-eyewear-glasses/\" target=\"_blank\" rel=\"noopener\">9to5Google: Google launches Intelligent Eyewear glasses (12 Nov 2026)</a></li>\n<li><a href=\"https://www.kunstpuntgroningen.nl/blog/statement-cultuursector-groningen-geen-slimme-brillen-in-onze-gebouwen-en-tijdens-onze-evenementen/\" target=\"_blank\" rel=\"noopener\">Kunstpunt Groningen: Statement Cultuursector Groningen (9 Oct 2026)</a></li>\n<li><a href=\"https://nltimes.nl/2026/10/09/35-groningen-clubs-museums-events-jointly-ban-smart-glasses-privacy-infringements\" target=\"_blank\" rel=\"noopener\">NL Times: 35 Groningen clubs, museums, events jointly ban smart glasses (9 Oct 2026)</a></li>\n<li><a href=\"https://www.oogtv.nl/2026/10/groningse-cultuursector-wil-slimme-brillen-met-camera-en-ai-weren-uit-gebouwen-en-bij-evenementen/\" target=\"_blank\" rel=\"noopener\">OOG TV: Groningse cultuursector wil slimme brillen weren (9 Oct 2026)</a></li>\n<li><a href=\"https://www.clickondetroit.com/tech/2026/10/05/norway-seeks-temporary-ban-on-smart-glasses-as-privacy-concerns-rise/\" target=\"_blank\" rel=\"noopener\">Associated Press via ClickOnDetroit: Norway seeks temporary ban on smart glasses (5 Oct 2026)</a></li>\n<li><a href=\"https://www.aol.co.uk/articles/piracy-fears-prompt-calls-ban-140632000.html\" target=\"_blank\" rel=\"noopener\">The Guardian via AOL: UK cinemas look at banning Meta smart glasses over piracy fears (20 Aug 2026)</a></li>\n<li><a href=\"https://vr.org/articles/uk-venues-ban-meta-smart-glasses-wetherspoons-soho-house-2026\" target=\"_blank\" rel=\"noopener\">VR.org: UK Pubs, Theatres, and Comic-Con Are Writing Meta&#39;s Glasses Out of the House Rules (10 Aug 2026)</a></li>\n<li><a href=\"https://vr.org/smart-glasses-bans\" target=\"_blank\" rel=\"noopener\">VR.org: Smart Glasses Bans 2026 tracker (updated 15 Nov 2026)</a></li>\n<li><a href=\"https://www.engadget.com/2245776/meta-closing-loophole-that-allowed-people-to-record-with-smart-glasses-light-covered/\" target=\"_blank\" rel=\"noopener\">Engadget: Meta is closing a loophole that allowed people to record with their smart glasses&#39; light covered (27 Aug 2026)</a></li>\n<li><a href=\"https://about.fb.com/news/2026/10/meta-milans-pinacoteca-di-brera-and-essilorluxottica-expand-access-to-art-through-ai-glasses/\" target=\"_blank\" rel=\"noopener\">Meta Newsroom: Meta, Milan&#39;s Pinacoteca di Brera, and EssilorLuxottica Expand Access to Art Through AI Glasses (14 Oct 2026)</a></li>\n<li><a href=\"https://mixed-news.com/en/meta-ai-glasses-pinacoteca-brera-sensing-art-eight-artworks/\" target=\"_blank\" rel=\"noopener\">MIXED: Meta&#39;s AI glasses now describe eight Brera masterpieces to blind visitors in Milan (17 Oct 2026)</a></li>\n<li><a href=\"https://www.aestumanda.com/opinion/2026/07/smart-glasses-backlash-disabled-people/\" target=\"_blank\" rel=\"noopener\">Colin Hughes, Aestumanda: Calling smart glasses &#39;predator technology&#39; ignores disabled people who rely on them (30 Jul 2026)</a></li>\n<li><a href=\"https://press.epson.eu/en_GB/case-studies/national-theatre-uses-moverio-to-provide-captioning-for-hearing-impaired/\" target=\"_blank\" rel=\"noopener\">Epson: National Theatre uses Moverio to provide captioning (22 Feb 2019)</a></li>\n<li><a href=\"https://ningbo.chinadaily.com.cn/2026-08/06/c_1202944.htm\" target=\"_blank\" rel=\"noopener\">China Daily: AI glasses bring smart tour experiences to Tianyi Pavilion (6 Aug 2026)</a></li>\n</ul>\n",
      "image": "https://www.bcvr.io/blog/images/the-house-pair/hero-1200x630.jpg",
      "date_published": "2026-11-18T12:00:00.000Z",
      "date_modified": "2026-11-18T12:00:00.000Z",
      "tags": [
        "XR",
        "Installations",
        "Industry"
      ],
      "authors": [
        {
          "name": "Scott Marshall"
        }
      ]
    },
    {
      "id": "https://www.bcvr.io/blog/two-channels-wide/",
      "url": "https://www.bcvr.io/blog/two-channels-wide/",
      "title": "Two Channels Wide",
      "summary": "Generative video finally learned to make sound. It makes it already mixed, and in the same fortnight both dominant post applications shipped tools for pulling finished mixes apart.",
      "content_html": "<p>Adobe spent MAX, in Miami Beach from 10 to 12 November, putting generation inside the edit. Its new Generative Media tool drops Generate Video, Generate Sound Effects, Generate Soundscape and Generate Music straight into the Premiere timeline. In the same announcement, with far less fanfare, came tools doing the opposite. Enhance Audio &quot;allows you to separate audio sources and then mix them independently&quot;. Separate Crosstalk splits overlapping speakers onto their own tracks. Of ten research previews at Sneaks, exactly one concerned sound, and Project Sound Bender opens with subtraction too: describe a sound, pull it out of a finished mix, then generate variations.</p>\n<p>Days earlier, Avid shipped Pro Tools 2026.11 with a Stem Separator Extension. Drag in a file, get back up to six stems, vocals, drums, bass, guitar, keys and other, processed on your own machine.</p>\n<p>Separation itself is old news. Spleeter was open sourced in 2019 and Logic Pro has had a Stem Splitter since 2023. What changed this month is placement. Un-mixing became standard equipment in both applications most finished video passes through, in the same release cycle as generated audio.</p>\n<h2 id=\"generated-sound-arrives-welded\">Generated sound arrives welded</h2>\n<p>For anyone working in a room rather than on a timeline, the relationship between those two halves is the whole story.</p>\n<p>A mix is a decision about a listening position, and generated audio arrives with that decision already taken, because the finished object is the only thing the models were ever asked for. Google&#39;s model page sells Veo 3 on &quot;sound effects, ambient noise, and even dialogue&quot;, all of it generated natively. The public pages discuss audio as a feature. What they do not discuss, anywhere I could find, is audio as a <em>format</em>: no channel layout, no sample rate, no bed and objects structure. On a timeline that is a non-question. In a building it is the only question.</p>\n<figure class=\"post-figure\"><img src=\"https://www.bcvr.io/blog/images/two-channels-wide/figure-1.jpg\" alt=\"A beam of light passes through a glass block in an anechoic chamber and splits into coloured ribbons, each ending at a different loudspeaker\" width=\"1200\" height=\"670\" loading=\"lazy\" decoding=\"async\"><figcaption>The operation the industry has started shipping: one finished signal, pulled back into separable parts and sent to specific positions. The hard half is the second one.</figcaption></figure>\n<p>The clearest account of why sits in Apple&#39;s own research. StereoFoley, by Tornike Karchkhadze of UC San Diego with six Apple colleagues during an internship there, generates &quot;spatially accurate stereo sound at 48 kHz&quot;, placing sounds left and right according to where the object sits in frame. In a listening test of 1,341 ratings from 131 qualified raters, the object aware model scored 3.46 on a five point scale, against 2.93 for the base model and 2.19 for MMAudio, the mono state of the art. One line further down, the <em>real</em> recordings scored 2.97. A generated mix outscoring the original is a result about preference, not fidelity to a place.</p>\n<blockquote class=\"pull\"><p>A fixed mix can fill a room. It cannot be re-aimed in one, and generated sound arrives with its sources already welded together.</p>\n</blockquote><p>Getting that far took a workaround that tells you where the field is. VGGSound, its standard audio video training set, proved near useless for spatial learning. About 27% of clips are &quot;effectively mono&quot;, their channels differing on average by less than one part in a hundred thousand, and the authors estimate under 2% of the corpus shows clear object aware stereo placement at all. So they synthesised the spatial training data instead, captioning scenes with Gemini, segmenting objects, generating generic sounds, panning them by rule. Their conclusion runs to a line: &quot;the main challenge for object-aware stereo audio generation is not architectural but data-related.&quot;</p>\n<p>The evaluation carries a second confession. Standard metrics are built on mono backbones, so stereo outputs &quot;must be resampled and downmixed to mono, making the metrics largely insensitive to high-frequency content and stereo characteristics.&quot; The field cannot yet reliably measure the spatial quality of what it makes.</p>\n<h2 id=\"spatial-generation-exists-one-rung-up\">Spatial generation exists, one rung up</h2>\n<p>The exception shows the shape of the problem rather than solving it. ImmerseDiffusion, a 2024 paper from the same Apple group led by Mojtaba Heydari, then at the University of Rochester, generates first order ambisonics: four channels, steerable by description or parametrically, through azimuth, elevation, distance, room size and decay time. Its training data was itself synthetically spatialised, thousands of hours built mostly from an existing Freesound corpus. Its best azimuth error is 0.36 against a 0.17 floor the paper calls the lower bound a model could reach. No unit is given, though the circular difference formula implies radians, about 21 degrees if so. First order is the bottom rung of spatial resolution in any case. The paper &quot;focuses exclusively on generating FOA components&quot;, generalisable upward &quot;albeit with increased complexity&quot;. Four channels around one listening point is a fine deliverable for headphones, not a show file for forty loudspeakers.</p>\n<h2 id=\"one-head-in-a-known-position\">One head, in a known position</h2>\n<p>The most complete attempt to join the halves appeared at UIST in Detroit this month. Sonic Stage, from a team across HKUST, Columbia, Stuttgart and Rochester, reconstructs a dialogue scene in 3D, works out which sounds are missing, generates five candidates each with ElevenLabs&#39; text to sound model, has Gemini rate them, and places the winners in the reconstructed space. The gains are not marginal. Blind and low vision viewers recalled character positions 89% of the time against 44% for the baseline, and movement 86% against 18%.</p>\n<p>The honest numbers are the useful ones. Twelve participants, on AirPods Pro, 8.2 minutes of processing per minute of video. The lowest rated of the study&#39;s audio measures was diegetic clarity, whether you could tell <em>what</em> the generated sound was, at 5.00 out of 7, standard deviation 1.28 across those twelve people. The system &quot;assumes non-overlapping speech&quot;, and the authors note that &quot;when several actions occur in rapid succession or simultaneously, auditory cues may become difficult to distinguish.&quot; Generated spatial audio presumes one head whose position is known. A gallery has neither.</p>\n<p>Vendors are edging toward rooms regardless. ElevenLabs made its third generation sound effects model the default across its creative tools on 2 November, with a flag for effects that loop smoothly, and a looping effect is an installation feature rather than an advertising one. A week later its Unity agents SDK arrived, playing a voice through an AudioSource you position in the scene &quot;for spatial audio&quot;. The spatialisation is delegated to the game engine, as it usually is.</p>\n<p>In fairness, separating sound by position is not unsolved in general. Beamforming and direction of arrival separation are decades old, and Zylia&#39;s ambisonic Beamformer plugin costs 149 euros. But those tools separate a <em>captured</em> sound field, recorded by an array that knew where it was. They do nothing for a two channel file a model made ten seconds ago. Pro Tools has carried MPEG-H mixing with an Immersive Panner since April 2026. This month it got a separator. The separator hands you a guitar; the panner wants a coordinate. Nothing joins them, and until something does, the join is a person.</p>\n<figure class=\"post-figure\"><img src=\"https://www.bcvr.io/blog/images/two-channels-wide/figure-2.jpg\" alt=\"Title card reading Two Channels Wide, The Immersive Dispatch\" width=\"1200\" height=\"679\" loading=\"lazy\" decoding=\"async\"><figcaption>The Immersive Dispatch, 17 November 2026.</figcaption></figure>\n<h2 id=\"what-it-costs-on-a-real-job\">What it costs on a real job</h2>\n<p>At BCVR, my studio, the reading is unromantic. Generative tools collapsed the cost of material and left the cost of placement where it was. On a hybrid physical and digital install the artefact we invoice is not the sound. It is a positioned cue list that survives a venue change, degrades gracefully when a driver fails at 11am on a Saturday, and can be re-aimed in a room whose drawings were wrong. Generation saves the library search, not the mix. Budget the de-mix rather than the generate, and the estimate stops drifting.</p>\n<p>Which points at the asset hardly anyone is making. If under 2% of the field&#39;s standard training set holds cleanly placed sources, then dry, single source, position logged recordings of real things in real rooms are at once show grade and training grade material, and they are thrown away daily. Record dry, log the position, keep the stems. A mix can always be made later. It can never be unmade.</p>\n<h2 id=\"sources\">Sources</h2>\n<ol>\n<li><a href=\"https://blog.adobe.com/en/publish/2026/11/10/explore-new-ways-create-across-creative-cloud-adobe-firefly-while-advancing-craft-precision-choice-control\" target=\"_blank\" rel=\"noopener\">Adobe Blog, Creative Cloud and Firefly announcements, 10 November 2026</a></li>\n<li><a href=\"https://blog.adobe.com/en/publish/2026/11/11/adobe-max-sneaks\" target=\"_blank\" rel=\"noopener\">Adobe MAX Sneaks 2026, 11 November 2026</a></li>\n<li><a href=\"https://www.tvtechnology.com/platform/broadcast/pro-tools-adds-new-ai-powered-extension-for-separating-audio-into-individual-elements\" target=\"_blank\" rel=\"noopener\">Jenny Priestley, Pro Tools Stem Separator Extension, TVTechnology, 6 November 2026</a></li>\n<li><a href=\"https://www.production-expert.com/production-expert-1/pro-tools-2026-11-adds-extensions-built-in-stem-separation-and-mpe\" target=\"_blank\" rel=\"noopener\">Production Expert, Pro Tools 2026.11, 3 November 2026</a></li>\n<li><a href=\"https://www.avid.com/pro-tools/whats-new\" target=\"_blank\" rel=\"noopener\">Avid, What&#39;s New in Pro Tools</a></li>\n<li><a href=\"https://deepmind.google/models/veo/\" target=\"_blank\" rel=\"noopener\">Google DeepMind, Veo model page</a></li>\n<li><a href=\"https://arxiv.org/abs/2509.18272\" target=\"_blank\" rel=\"noopener\">StereoFoley, arXiv:2509.18272</a></li>\n<li><a href=\"https://arxiv.org/abs/2410.14945\" target=\"_blank\" rel=\"noopener\">ImmerseDiffusion, arXiv:2410.14945</a></li>\n<li><a href=\"https://arxiv.org/abs/2607.20835\" target=\"_blank\" rel=\"noopener\">Sonic Stage, UIST 2026, arXiv:2607.20835</a></li>\n<li><a href=\"https://elevenlabs.io/docs/changelog/2026/11/2\" target=\"_blank\" rel=\"noopener\">ElevenLabs changelog, 2 November 2026</a></li>\n<li><a href=\"https://elevenlabs.io/docs/changelog/2026/11/9\" target=\"_blank\" rel=\"noopener\">ElevenLabs changelog, 9 November 2026</a></li>\n<li><a href=\"https://www.zylia.co/shop/zylia-beamformer-543\" target=\"_blank\" rel=\"noopener\">Zylia Beamformer plugin</a></li>\n</ol>\n",
      "image": "https://www.bcvr.io/blog/images/two-channels-wide/hero-1200x630.jpg",
      "date_published": "2026-11-17T12:00:00.000Z",
      "date_modified": "2026-11-17T12:00:00.000Z",
      "tags": [
        "Spatial sound",
        "AI film",
        "Industry"
      ],
      "authors": [
        {
          "name": "Scott Marshall"
        }
      ]
    },
    {
      "id": "https://www.bcvr.io/blog/nobody-budgets-for-ram/",
      "url": "https://www.bcvr.io/blog/nobody-budgets-for-ram/",
      "title": "Nobody Budgets for RAM",
      "summary": "Every XR price forecast assumes someone absorbs component inflation, and the two forecasters you would ask disagree by two years.",
      "content_html": "<p>Every price forecast in XR assumes somebody absorbs component inflation. That assumption is now load-bearing, and the two people you would ask about it disagree by two years.</p>\n<p>On 16 April, Meta published a short note about Quest pricing. Three days later the Quest 3S went from $299.99 to $349.99 and the 512GB Quest 3 from $499.99 to $599.99. &quot;The cost of building high-performance VR hardware has risen significantly,&quot; the company wrote, and pointed at memory chips.</p>\n<p>Valve had got there first. In February, having announced the Steam Frame and Steam Machine the previous November without prices, it said it needed to revisit both the shipping schedule and the pricing, because &quot;the memory and storage shortages you&#39;ve likely heard about across the industry have rapidly increased since then.&quot;</p>\n<p>Two device makers repricing on components inside two months is unusual. What turns it into a planning problem is that nobody agrees when it stops.</p>\n<h2 id=\"the-two-year-gap\">The two-year gap</h2>\n<p>IDC&#39;s memory crisis note, published on 26 August, says memory prices are &quot;expected to continue increasing until at least 2028&quot; and describes the market that emerges &quot;on the other side of the crisis, when the memory supply finally stabilizes in 2028.&quot;</p>\n<p>Two days before that, at the groundbreaking for SK hynix&#39;s packaging plant in Indiana, chief executive Kwak Noh-jung forecast a global memory chip shortage through the end of 2030, with supply and demand balancing after that. In June the company&#39;s chairman, Chey Tae-won, put it more mildly: &quot;Until 2030, there&#39;s still some shortage.&quot; SK hynix is doubling its memory wafer capacity within five years, but Chey put the lead time for a greenfield fab at more than five years, so on his own numbers the new output lands about when the squeeze is meant to end anyway.</p>\n<p>Take the second forecast with the scepticism it deserves. SK hynix is one of the three firms that would have to supply any normalisation, and its margins improve the longer the shortage runs. It is also the company that would know. Somewhere between 2028 and 2030 is where every hardware line in every immersive proposal written this year resolves.</p>\n<h2 id=\"what-the-asp-forecast-is-actually-saying\">What the ASP forecast is actually saying</h2>\n<p>IDC&#39;s smart glasses outlook, published in June, has display-less glasses at 13.6 million units and $5.1bn this year, rising to 27.3 million units by 2030, with the average selling price falling from $376 to $229. For the least memory-hungry product in the category, that ordinary volume-up, price-down curve is probably right.</p>\n<p>The mixed reality line is more interesting. ASP goes from $742 in 2026 to $682 in 2030: an eight per cent decline over four years, flat in nominal terms and rising in real ones, in a category that has historically deflated far faster. A flat ASP forecast is what sustained cost pressure looks like on a spreadsheet.</p>\n<p>The outlook does not mention memory, DRAM, components or supply anywhere in it. What it does say is that falling ASPs drive volume, but that they also mean &quot;vendors who compete purely on hardware will face margin pressure, and software, services, and AI differentiation will become the real moat.&quot;</p>\n<p>Read that next to Meta&#39;s April note and the two stop contradicting each other. Prices fall to the buyer because vendors eat the difference. Meta&#39;s price rise is a vendor declining to keep eating it.</p>\n<h2 id=\"the-demand-side-is-not-the-problem\">The demand side is not the problem</h2>\n<p>In the first quarter of 2026, display-less smart glasses shipped about 2.25 million units, up 167 per cent year on year, a single quarter close to the whole of 2024. Display eyewear grew 86 per cent. Meta held 69.2 per cent of the smart glasses market overall.</p>\n<p>The component numbers underneath are the ones to watch. TrendForce put conventional DRAM contract prices up by roughly 93 to 98 per cent quarter on quarter in the first quarter of this year, with memory industry revenue up 81 per cent to $97bn in that quarter alone. Increases are moderating into the third quarter, but from a very high base, and TrendForce is blunt about where that lands: &quot;record-high contract prices mean customers from consumer markets, such as PCs and smartphones, are reaching their affordability limit.&quot;</p>\n<p>TrendForce reported in January that the waveguide order for Meta&#39;s Ray-Ban Display had been revised up from around 80,000 units to 150,000 within six months, and forecast 950,000 AR glasses shipments for 2026, up 53 per cent, naming capacity limits in optics, displays and assembly as the constraints on scaling. Samsung and Google&#39;s Android XR glasses are teased for this autumn, on a timeline last confirmed in May, with display versions due in 2027. Nobody in this market is short of buyers. They are short of parts.</p>\n<h2 id=\"what-a-studio-does-when-the-device-gets-expensive\">What a studio does when the device gets expensive</h2>\n<p>On 28 August, Schell Games cut 10 per cent of its staff, leaving around 125 people. &quot;In 24 years as a studio, this is the first time we&#39;ve found it necessary to have a layoff,&quot; Jesse Schell said, describing a move toward &quot;a greater number of smaller projects&quot; and citing &quot;the way the industry is right now.&quot; Road to VR supplies context Schell did not: Meta&#39;s studio closures and cancellations this year shifted the VR funding environment.</p>\n<p>He did not blame memory prices and neither will I. But keep one number. VR is now about 30 per cent of the studio&#39;s work, down from around 80 per cent, and the growth area is what it calls VR-adjacent work, including interactive projection for location-based entertainment. A studio that spent two decades building for headsets is moving toward things you can walk into without one.</p>\n<h2 id=\"from-where-i-sit\">From where I sit</h2>\n<p>When a client asks for the hardware line in a proposal, the question underneath it is cost per visitor. A headset build puts a computer in every pair of hands, and with it comes capex, charging, breakage, staff to fit and sanitise, and a refresh cycle quoted at a price you cannot hold. A projected or LED build puts fewer, larger machines in a room. Those are not memory-free, and media servers have their own supply stories, but they carry far less memory per head of audience served. That is my read, not a published figure. Both can be the right answer. The difference now is that the headset option carries a line I cannot forecast ninety days out.</p>\n<p>The generative side cuts the other way, and it is the one piece of good news in the arithmetic. Content is the half getting cheaper: variants, previz and language versions have all come down in real terms for us over the past two years, and that saving goes into the scope rather than the margin.</p>\n<blockquote class=\"pull\"><p>Content is deflating and the devices are not, so the contingency belongs on the side with a chip in it.</p>\n</blockquote><h2 id=\"what-to-do-about-it\">What to do about it</h2>\n<p>Cost per visitor sets throughput, and throughput is the number a brand reports on afterwards, so a device-per-person design costed in 2024 is worth re-running before it reaches a purchase order. Ask suppliers to quote hardware with a validity window written on it, and treat anything past ninety days as an estimate. If a project has a phase two in 2027, either buy the devices now or design phase two so it does not need any. And when a forecast says the glasses will be cheap by then, find out who is assumed to be absorbing the difference, because at the moment it is the vendor, and vendors stop.</p>\n<p>We have started putting a dated line in our own quotes saying which week the hardware pricing came from. It looks fussy on the page. It has already saved one awkward conversation.</p>\n<p>Scott Marshall runs BCVR, a London studio building immersive and mixed reality installations for brands and cultural venues. <a href=\"https://www.bcvr.io\">https://www.bcvr.io</a></p>\n<h2 id=\"sources\">Sources</h2>\n<ol>\n<li><a href=\"https://www.meta.com/blog/update-meta-quest-pricing/\" target=\"_blank\" rel=\"noopener\">Meta, &quot;An Update on Meta Quest Pricing&quot;, 16 April 2026</a></li>\n<li><a href=\"https://roadtovr.com/meta-raises-quest-3-3s-prices-memory-shortage/\" target=\"_blank\" rel=\"noopener\">Road to VR, &quot;Meta is Raising Quest 3 and Quest 3S Prices Amid Memory Cost Surge&quot;, 16 April 2026</a></li>\n<li><a href=\"https://roadtovr.com/steam-frame-ram-shortage-release-date-price/\" target=\"_blank\" rel=\"noopener\">Road to VR, &quot;Valve Reconsiders Steam Frame Price &amp; Release Date Amid RAM &amp; Storage Shortage&quot;, 5 February 2026</a></li>\n<li><a href=\"https://www.koreaherald.com/article/10855627\" target=\"_blank\" rel=\"noopener\">The Korea Herald, &quot;SK hynix open to US investment, sees memory shortage through 2030&quot;, 28 August 2026</a></li>\n<li><a href=\"https://www.tomshardware.com/pc-components/dram/sk-hynix-to-double-memory-wafer-capacity-over-five-years\" target=\"_blank\" rel=\"noopener\">Tom&#39;s Hardware, &quot;SK hynix to double memory wafer capacity within five years, chairman says&quot;, 2 June 2026</a></li>\n<li><a href=\"https://www.idc.com/resource-center/blog/smart-glasses-surge-the-xr-market-is-rewriting-its-own-rules/\" target=\"_blank\" rel=\"noopener\">IDC, &quot;Smart Glasses Market 2026: XR Is Rewriting the Rules&quot;, 15 June 2026</a></li>\n<li><a href=\"https://www.idc.com/resource-center/blog/smartphone-shipments-set-for-record-16-7-drop-in-2026-as-the-memory-crisis-hits-full-force/\" target=\"_blank\" rel=\"noopener\">IDC, &quot;Smartphone Shipments Set for Record 16.7% Drop in 2026, as the Memory Crisis Hits Full Force&quot;, 26 August 2026</a></li>\n<li><a href=\"https://www.trendforce.com/presscenter/news/20260601-13070.html\" target=\"_blank\" rel=\"noopener\">TrendForce, &quot;Rapid Contract Price Surge Drives 1Q26 DRAM Industry Up 81% QoQ&quot;, 1 June 2026</a></li>\n<li><a href=\"https://www.trendforce.com/presscenter/news/20260703-13134.html\" target=\"_blank\" rel=\"noopener\">TrendForce, &quot;AI Server Demand Continues to Support Memory Prices in 3Q26&quot;, 3 July 2026</a></li>\n<li><a href=\"https://www.trendforce.com/presscenter/news/20260130-12908.html\" target=\"_blank\" rel=\"noopener\">TrendForce, &quot;Component Orders for Meta Ray-Ban Display Glasses Revised Up Twice&quot;, 30 January 2026</a></li>\n<li><a href=\"https://roadtovr.com/among-us-vr-studio-schell-games-layoffs-aug-2026/\" target=\"_blank\" rel=\"noopener\">Road to VR, &quot;Veteran VR Studio Schell Games Confirms Layoffs Affecting 10 Percent of Staff&quot;, 28 August 2026</a></li>\n<li><a href=\"https://9to5google.com/2026/05/19/google-samsung-android-xr-glasses-styles-release-date/\" target=\"_blank\" rel=\"noopener\">9to5Google, &quot;Google and Samsung&#39;s Android XR glasses launch in &#39;Fall&#39; 2026&quot;, 19 May 2026</a></li>\n</ol>\n",
      "image": "https://www.bcvr.io/blog/images/nobody-budgets-for-ram/hero-1200x630.jpg",
      "date_published": "2026-09-01T12:00:00.000Z",
      "date_modified": "2026-09-01T12:00:00.000Z",
      "tags": [
        "Industry",
        "XR",
        "Installations"
      ],
      "authors": [
        {
          "name": "Scott Marshall"
        }
      ]
    },
    {
      "id": "https://www.bcvr.io/blog/the-scan-is-the-prompt/",
      "url": "https://www.bcvr.io/blog/the-scan-is-the-prompt/",
      "title": "The Scan Is the Prompt",
      "summary": "World models will invent everything you do not give them. On an installation, the one thing you cannot let them invent is the building.",
      "content_html": "<p>World Labs released Marble last November: explorable 3D worlds built from text, images, video or coarse 3D layouts, exported as Gaussian splats, meshes or video, and the first generative world tool many production teams could plausibly put in a pipeline. The line worth reading twice is not a boast but a description of how prompting works. Text and image prompts, the company writes, are &quot;intuitive and powerful, but limited in creative control: Marble must invent all the details of the world that are not present in the input text or image prompt.&quot; Often magical, it adds; sometimes not what you want, which is precisely why the product also accepts a coarse 3D layout as an input.</p>\n<p>That last option is the argument of this piece, arriving early and from the vendor: the way you stop a world model inventing your building is to hand it the building.</p>\n<p>Google DeepMind is equally explicit about Genie 3, its navigable world model running at 720p and 24 frames per second. Its published limitations include interaction duration (&quot;a few minutes of continuous interaction, rather than extended hours&quot;), environments &quot;largely consistent for several minutes&quot; with visual memory reaching about a minute back, and an inability to &quot;simulate real-world locations with perfect geographic accuracy.&quot;</p>\n<p>Set that against a typical installation brief: one specific real building, at exact geometry, running for months. Generative models are fluent about surfaces and agnostic about coordinates. The question was never whether AI can generate your content (it plainly can) but what you feed it so it stops inventing your venue.</p>\n<p>That is the job a digital twin has quietly taken on. Not visualisation. Ground truth.</p>\n<h2 id=\"what-the-model-will-not-give-you\">What the model will not give you</h2>\n<p>Three things a generative system will not supply for a physical build, and a survey will.</p>\n<p>The first is geographic truth. DeepMind&#39;s caveat is the tell: the model produces a convincing room, but not <em>your</em> room. A projection surface that is convincingly wrong is worse than a blank one, because it will not line up with a column that exists.</p>\n<p>The second is repeatability, and it is worth being precise. For most installations the model&#39;s session ceiling is beside the point: content is rendered once and played back. The real constraint is that generative output is not reproducible on demand. Ask in month four for the same shot with one element changed, and a pipeline that cannot regenerate against fixed geometry returns something subtly different. On a projection surface, that means a realignment. Venues running models live inherit the session limit directly, and it is short.</p>\n<p>The third is tolerance, and here I am reporting from practice rather than citing anyone, because it is not the kind of thing a model card measures. A generated frame carries no obligation to be right to within a few millimetres of a real edge. A projection map does. Every frame in an installation is ultimately a claim about where a physical surface is.</p>\n<blockquote class=\"pull\"><p>A generative model is fluent about surfaces and completely agnostic about coordinates. A building is a coordinate problem.</p>\n</blockquote><h2 id=\"the-capture-layer-grew-up\">The capture layer grew up</h2>\n<p>The useful development of the last eighteen months is not generative at all. It is that measuring a room got cheap.</p>\n<p>Gaussian splatting is the reason. AWE&#39;s own write-up ahead of its 2026 US event called it &quot;the biggest story in reality capture,&quot; noting that a replica of a complex space that once &quot;took weeks of processing and massive computing power&quot; can now be captured in startling detail and viewed in real time on standard hardware. The conference put a Digital Twin track and a Reality Capture Pavilion on the floor and added an Auggie award category for reality capture: the unglamorous signal that a technology has become infrastructure. The 3D mapping and modelling market is forecast to run from $17.63bn in 2026 to $29.20bn by 2031, a 10.6% compound rate, with mapping hardware still the largest slice.</p>\n<p>It is worth being honest about what splats are not. The Portuguese commercial capture studio Amplified Creations (which sells this work, and whose recommended hybrid workflow happens to be the one argued for below) puts it well in a field report: &quot;a splat is a cloud of colour, not a structured model.&quot; Excellent for looking at, poor for measuring, heavy to deliver, largely uneditable after capture, and unforgiving of bad input, which leaves floaters that are hard to fix afterwards. Which is why the serious version of this work is hybrid: laser survey for the geometry you build against, photogrammetry or splats for appearance, and the discipline to know which you are trusting for which decision.</p>\n<p>The most complete example remains the Vatican&#39;s. Iconem spent three weeks scanning St Peter&#39;s Basilica with drones, cameras and lasers, gathering more than 400,000 high-resolution images; Microsoft&#39;s AI for Good Lab refined the result to millimetre-level accuracy, then ran AI over it to detect cracks and missing mosaic tiles. That twin became two immersive public exhibits and an interactive site, launched in November 2024 ahead of the Jubilee. Note the order of operations: the building was measured first, and the AI worked on the measurement.</p>\n<h2 id=\"the-survey-is-the-first-creative-act\">The survey is the first creative act</h2>\n<p>At BCVR the change we have felt most is not in the render farm. It is that the site survey stopped being a technical formality and became the moment the project actually gets decided.</p>\n<p>Once you hold a good capture, a surprising number of arguments resolve themselves. Sightlines stop being opinion. You find out in week one, not on site, that the column nobody mentioned bisects the hero shot. Content is authored against the real surface rather than a drawing of it: the difference between a two-day alignment on site and a two-week one. And the same capture that carries the projector map also carries the acoustic model: if audio placement is where the emotional work happens, one measured model of the room feeding both visual and sonic design is the cheapest coordination a project will ever buy.</p>\n<p>There is a governance benefit too, dull until it saves you. A generated asset authored against a surveyed twin has a provenance trail: made for these coordinates, from this capture, on this date. When a client&#39;s brand-safety review asks what the model was given, &quot;the building&quot; is a better answer than &quot;a prompt.&quot;</p>\n<h2 id=\"capture-first-then-generate\">Capture first, then generate</h2>\n<p>The practical read here is an order-of-operations argument rather than a technology one.</p>\n<p>Scan before you write the creative treatment, not after it is approved. Budget capture as a design cost rather than a survey line, because it is now the input that determines what the generative stage can be trusted to do. Decide which artefact is your measurement of record and which is your reference for appearance, and never let a splat be both. And when a world model offers to build the room for you, remember what its own documentation says it does with everything you did not give it.</p>\n<p>The models have become very good at imagining spaces. The advantage belongs to whoever has actually measured one.</p>\n<p><em>Scott Marshall is founder of BCVR (<a href=\"https://www.bcvr.io\">https://www.bcvr.io</a>), a London-based immersive production studio working across XR, projection mapping, spatial audio and digital twins for brand and cultural clients.</em></p>\n<h2 id=\"sources\">Sources</h2>\n<ol>\n<li><a href=\"https://www.worldlabs.ai/blog/marble-world-model\" target=\"_blank\" rel=\"noopener\">World Labs, Marble, 12 November 2025</a></li>\n<li><a href=\"https://deepmind.google/discover/blog/genie-3-a-new-frontier-for-world-models/\" target=\"_blank\" rel=\"noopener\">Google DeepMind, &quot;Genie 3: A new frontier for world models&quot;, 5 August 2025</a></li>\n<li><a href=\"https://www.awexr.com/blog/digital-twins-and-reality-capture-at-AWE-USA-2026\" target=\"_blank\" rel=\"noopener\">AWE, &quot;The Spatial Enterprise: Digital Twins and Reality Capture at AWE USA 2026&quot;, 20 March 2026</a></li>\n<li><a href=\"https://www.globenewswire.com/news-release/2026/07/23/3332347/0/en/3d-mapping-and-modeling-market-surges-to-29-20-billion-at-a-cagr-10-6-by-2031-report-by-marketsandmarkets.html\" target=\"_blank\" rel=\"noopener\">MarketsandMarkets via GlobeNewswire, 23 July 2026</a></li>\n<li><a href=\"https://amplifiedcreations.com/journal/gaussian-splatting-production-ready-2026\" target=\"_blank\" rel=\"noopener\">Amplified Creations, &quot;Is 3D Gaussian splatting production-ready in 2026? An honest field report&quot;, 4 July 2026</a></li>\n<li><a href=\"https://unlocked.microsoft.com/vatican/\" target=\"_blank\" rel=\"noopener\">Microsoft Unlocked, &quot;La Basilica di San Pietro&quot;, November 2024</a></li>\n</ol>\n",
      "image": "https://www.bcvr.io/blog/images/the-scan-is-the-prompt/hero-1200x630.jpg",
      "date_published": "2026-08-31T12:00:00.000Z",
      "date_modified": "2026-08-31T12:00:00.000Z",
      "tags": [
        "Installations",
        "AI film",
        "Industry"
      ],
      "authors": [
        {
          "name": "Scott Marshall"
        }
      ]
    },
    {
      "id": "https://www.bcvr.io/blog/immersion-is-an-ear-problem/",
      "url": "https://www.bcvr.io/blog/immersion-is-an-ear-problem/",
      "title": "Immersion Is an Ear Problem",
      "summary": "Controlled studies keep finding that spatial audio buys presence more cheaply than the industry assumes. The budget templates still tend not to have a line for it.",
      "content_html": "<p>In 2022 a team publishing in <em>Frontiers in Signal Processing</em> did something the immersive industry mostly avoids: they varied picture and sound independently and asked which one people felt. Seventeen trained listeners watched the same scenes at three video resolutions (0.5, 1.5 and 2.5 megapixels per eye) under three audio conditions, from mono up to head-tracked binaural with room acoustic rendering.</p>\n<p>Adding room acoustics to head-tracked binaural produced the same gain in reported immersion as multiplying the pixel count by five. Full auralisation at 20% video scored 5.79; head-tracked binaural at 100% video scored 5.76: no statistically significant difference, p = 0.891.</p>\n<p>This is not a finding that sound beats pictures. Video resolution had its own significant effect on immersion, and a larger effect than audio on audio-visual quality judgements. The authors bound the comparison to what they tested: a headset, headphones, and seventeen trained listeners they suggest retesting on naïve subjects. Reading it across to a loudspeaker-based room is an inference, not their claim; it is a reasonable one, and it should have changed how rooms get costed.</p>\n<p>It has not. One widely circulated 2026 cost guide for interactive installations splits a project into hardware at 30 to 50%, software and content at 30 to 50%, and fabrication and installation at 15 to 30%. Audio is not a category. One studio&#39;s template is not an industry survey, but the omission will be familiar to anyone who has priced this work: audio sits unnamed inside someone else&#39;s percentage, which is how a discipline gets value-engineered out at the second budget review. The same studio&#39;s sound guide argues for 5 to 15% of project cost (its own estimate, not a benchmark), which tells you both that practitioners know it matters and that they are arguing for scraps.</p>\n<h2 id=\"what-the-body-does-before-the-visitor-decides\">What the body does before the visitor decides</h2>\n<p>The case for sound is not aesthetic. It is physiological, and it is measurable.</p>\n<p>A study in <em>Frontiers in Virtual Reality</em> in July 2025 put 68 participants through VR content with and without software spatial audio enhancement (dynamic gain shaping plus HRTF-based localisation) and analysed physiology from 58 of them. In the commercial title tested, self-reported audio quality, auditory involvement and sound localisation all rose significantly; in the second title only involvement did, which is the honest version of the result. The body moved too: facial EMG showed shifted valence, arousal rose significantly toward the end of a scripted helicopter event, and heart rate variability dropped significantly in the enhanced condition during that event&#39;s final ten seconds. A separate published dataset (31 participants, 24-channel EEG, wristband blood volume pressure and electrodermal activity) points the same way: earlier work on it reported heart rate variability tracking the <em>involvement</em> dimension of presence specifically.</p>\n<p>The practical claim is narrow but strong: audio changes autonomic state, on the timescale of an event rather than a visit. A visitor cannot tell you why a room got tense. Their heart rate can.</p>\n<blockquote class=\"pull\"><p>Adding room acoustics to head-tracked binaural moved immersion as much as multiplying the pixel count by five. One of those two upgrades is a software decision.</p>\n</blockquote><h2 id=\"the-frontier-is-control-not-volume\">The frontier is control, not volume</h2>\n<p>Almost none of the current toolset is about being louder.</p>\n<p>Sphere Immersive Sound, built on HOLOPLOT&#39;s wave field synthesis and 3D audio-beamforming, runs roughly 1,600 permanently installed X1 Matrix Array modules plus 300 mobile ones: 167,000 individually amplified drivers for audiences up to 20,000. What that buys is not level but consistency and address: audio that holds quality across distance, different content sent simultaneously to different sections of the same room, and a whisper that arrives as though the speaker were beside you.</p>\n<p>Below that scale, object-based systems are now standard specification. d&amp;b&#39;s Soundscape handles up to 128 sound objects across 64 loudspeaker outputs, with an acoustic-emulation module running 144 convolvers and impulse responses up to 6.3 seconds: an artificial room laid over a real one. L-Acoustics&#39; L-ISA Processor II carries 128 inputs and outputs at 96 kHz, and its DJ product uses machine-learning source separation to split a mix into four sources in real time for placement. Meyer&#39;s Constellation does the reverse trick electroacoustically, changing a room&#39;s apparent acoustics on a preset; Spacemap Go runs 32 inputs across 256 outputs from an iPad. At the other extreme, directional arrays put a narrow beam of audio over one exhibit and leave the next one silent.</p>\n<p>The frontier has moved from reproducing sound to controlling where it exists. That is a spatial design problem, which is why it belongs in the room&#39;s design phase and not a post-production week.</p>\n<p>DATALAND, Refik Anadol&#39;s AI arts museum, opened at The Grand LA on 20 June 2026 with 250 L-Acoustics loudspeakers (its Data Pavilion carrying separate bottom, surround and height layers plus twenty-five X6i units for personalised audio). Anadol&#39;s framing is the useful one: &quot;Sound is not merely an accompaniment to my visual installations; it is an intrinsic element, co-constituting the overall sensory experience.&quot;</p>\n<h2 id=\"the-cheapest-presence-per-pound-on-the-drawing\">The cheapest presence per pound on the drawing</h2>\n<p>At BCVR this is the argument we have most often, and almost never with the creative team. It is with the budget.</p>\n<p>The pattern is predictable. A project is scoped with a strong sonic idea; the projector count rises during technical design; audio absorbs the difference. What gets lost is rarely fidelity (cheap speakers sound fine in a noisy room). It is <em>placement</em>. Once a sound is no longer attached to a specific point in space, the room stops being a place and becomes a screen with a soundtrack. Every mechanism in the research above runs through localisation, and localisation is the first thing a compromised system gives up.</p>\n<p>The fixes are unglamorous. Decide where the listener&#39;s head will be before choosing loudspeakers: a system for a seated audience and one for people wandering are different systems. Put a number on the room&#39;s noise floor at design stage rather than discovering it on site. And give audio its own budget line, because a discipline without a line item funds everyone else&#39;s overruns. The return shows up where clients actually measure: dwell, completion, whether people go quiet. Vendor-cited figures putting audio at 30 to 40% of perceived immersion, or good sound adding double digits to dwell, are worth what second-hand citations on a supplier blog are worth; the peer-reviewed physiology is what to take into a client meeting.</p>\n<h2 id=\"put-a-line-in-the-budget\">Put a line in the budget</h2>\n<p>None of this argues for spending less on the visual. It argues the ratio is wrong, in a direction that is cheap to fix: room acoustic rendering, object placement and a specified noise floor are mostly design decisions and software, competing against pixels that cost money by the square metre.</p>\n<p>If you commission immersive work, ask for the audio design at the same review as the visual, and ask what happens to it when the budget moves. If you build it, price the listening position before the screen. The emotional payload of an immersive room has always been carried by sound. The paperwork should admit it.</p>\n<p><em>Scott Marshall is founder of BCVR (<a href=\"https://www.bcvr.io\">https://www.bcvr.io</a>), a London-based immersive production studio working across XR, projection mapping, spatial audio and digital twins for brand and cultural clients.</em></p>\n<h2 id=\"sources\">Sources</h2>\n<ol>\n<li><a href=\"https://www.frontiersin.org/journals/signal-processing/articles/10.3389/frsip.2022.904866/full\" target=\"_blank\" rel=\"noopener\">Frontiers in Signal Processing, &quot;On the Relative Importance of Visual and Spatial Audio Rendering on VR Immersion&quot; (2022)</a></li>\n<li><a href=\"https://www.frontiersin.org/journals/virtual-reality/articles/10.3389/frvir.2025.1629908/full\" target=\"_blank\" rel=\"noopener\">Frontiers in Virtual Reality, &quot;&#39;Did you hear that?&#39;: Software-based spatial audio enhancements increase self-reported and physiological indices on auditory presence and affect in virtual reality&quot;, 31 July 2025</a></li>\n<li><a href=\"https://link.springer.com/article/10.1186/s13104-024-07009-x\" target=\"_blank\" rel=\"noopener\">BMC Research Notes data note, &quot;Data on neurophysiological and psychological responses to audio immersive experience in stereo and 3D audio formats&quot;</a></li>\n<li><a href=\"https://www.sphereentertainmentco.com/sphere-entertainment-unveils-the-most-advanced-concert-grade-audio-system-in-the-world-sphere-immersive-sound-powered-by-holoplot/\" target=\"_blank\" rel=\"noopener\">Sphere Entertainment Co., Sphere Immersive Sound powered by HOLOPLOT</a></li>\n<li><a href=\"https://www.avnation.tv/2026/06/26/dataland-opens-as-the-worlds-first-museum-of-ai-arts-featuring-immersive-audio-by-l-acoustics-l-isa-and-ambiance-technologies/\" target=\"_blank\" rel=\"noopener\">AVNation, 26 June 2026</a></li>\n<li><a href=\"https://www.svconline.com/products/audio/more-0/tech-roundup-electroacoustics-and-spatial-audio-2026\" target=\"_blank\" rel=\"noopener\">Sound &amp; Video Contractor, &quot;Tech Roundup: Electroacoustics and Spatial Audio 2026&quot;, 6 June 2026, updated 18 August 2026</a></li>\n<li><a href=\"https://www.l-acoustics.com/press-releases/l-acoustics-pushes-the-boundaries-of-immerisve-audio-at-ise-2026/\" target=\"_blank\" rel=\"noopener\">L-Acoustics at ISE 2026 (3 to 6 February 2026, Barcelona)</a></li>\n<li><a href=\"https://www.utsubo.com/blog/interactive-installation-cost-budget-guide\" target=\"_blank\" rel=\"noopener\">Utsubo, &quot;Interactive Installation Cost Guide 2026&quot;</a></li>\n<li><a href=\"https://www.utsubo.com/blog/sound-design-interactive-installations-guide\" target=\"_blank\" rel=\"noopener\">Utsubo, &quot;Sound Design for Interactive Installations: 2026 Guide&quot;</a></li>\n<li><a href=\"https://www.holosonics.com/applications/museumsandgalleries\" target=\"_blank\" rel=\"noopener\">Holosonics, &quot;Museums and Galleries&quot;</a></li>\n</ol>\n",
      "image": "https://www.bcvr.io/blog/images/immersion-is-an-ear-problem/hero-1200x630.jpg",
      "date_published": "2026-08-28T12:00:00.000Z",
      "date_modified": "2026-08-28T12:00:00.000Z",
      "tags": [
        "Spatial sound",
        "Immersive",
        "Installations",
        "XR"
      ],
      "authors": [
        {
          "name": "Scott Marshall"
        }
      ]
    },
    {
      "id": "https://www.bcvr.io/blog/the-clip-is-free/",
      "url": "https://www.bcvr.io/blog/the-clip-is-free/",
      "title": "The Clip Is Free, the Minute Costs a Million",
      "summary": "OpenAI switches off the Sora API next month. Google's video models run a 16K screen that has taken over $400m. Generative video did not fail or win: it stopped being a product.",
      "content_html": "<p>On 24 September, OpenAI turns off the Sora API. The consumer app went dark on 26 April, a month after the company announced the wind-down on 25 March, and the help-centre notice reads like a decommissioning order: export your content, because everything associated with your use of Sora will be permanently deleted. The product that made text-to-video a household idea in February 2024 had, by the time it was cancelled, fallen from first to 126th on the App Store&#39;s free chart, and Disney had walked away from a three-year licensing deal (announced barely three months earlier, and reported to include a planned $1bn stake) that would have let more than 200 of its characters be generated on demand.</p>\n<p>Six weeks before that API shutdown date, Higgsfield closed a $400m Series B at a $5.4bn valuation, roughly four times the $1.3bn it was worth in January, on $700m of annualised revenue, 30 million users across 238 countries and 390 Fortune 500 customers. Both of these things are true, and neither is a contradiction. The consumer destination for generated video is closing. The industrial supply of it is compounding.</p>\n<p>What has actually happened is that generative video stopped being somewhere you go and became a stage in a pipeline. For anyone who builds physical experiences, that reclassification is the whole story.</p>\n<h2 id=\"the-arithmetic-of-a-minute\">The arithmetic of a minute</h2>\n<p>Put the two ends of the market next to each other.</p>\n<p>At the model end: Veo 3.1 generates in eight-second clips at 1080p and 4K, chainable to something over 140 seconds, and is listed by API resellers at roughly $0.40 per second with synchronised audio, $0.20 without. Call it $24 of raw generation for a minute of moving image with sound. A decade ago that minute was a shoot.</p>\n<p>At the venue end: <em>The Wizard of Oz at Sphere</em> runs 75 minutes on a 160,000-square-foot, 16K-by-16K screen, built by taking a 1939 Technicolor film in a four-by-three frame and using custom-tuned Veo, Imagen and Gemini models for super-resolution, outpainting the edges of shots that were never composed to be seen whole, and generating performances for actors the original camera had panned away from. Around 1.2 petabytes of data went through it. Thousands of researchers, engineers, VFX artists, archivists and producers were involved across Google DeepMind, Google Cloud, Sphere Studios, Magnopus and Warner Bros. Discovery. The New York Post reported a budget of $80m in 2024, when the show was still an 80-minute plan; later accounts put it nearer $100m. Against the 75 minutes that actually opened, that is somewhere north of a million dollars a minute.</p>\n<p>Same technology family. On the $80m figure, about forty-four thousand times the cost per minute.</p>\n<p>It is not a like-for-like comparison ($24 buys model inference, $80m buys an entire production), and the mismatch is the thing worth measuring. The gap is a precise inventory of everything the model does not do: hold a 16K frame together across a curved surface, survive colour management across thousands of LED panels, sit on the right side of a rights agreement with Warner Bros., and still be there, frame-accurate, on the eleventh show of the week. Buzz Hayes, Google Cloud&#39;s global lead for entertainment industry solutions, described the work as &quot;an epic production undertaking, achieved with still new and largely experimental technology&quot;. That is an unusually honest way of saying the generation was the part they could buy.</p>\n<blockquote class=\"pull\"><p>The model is now the cheapest thing in the room. Everything that makes its output survive contact with a real wall is not.</p>\n</blockquote><h2 id=\"the-feed-is-demoting-it-the-box-office-is-paying-for-it\">The feed is demoting it. The box office is paying for it.</h2>\n<p>The consumer numbers explain why the app model broke. Animoto&#39;s January report (450-plus respondents, surveyed the previous September) found 83% of consumers say they have watched a video they suspected was AI-generated, and 36% say AI-generated video lowers their perception of the brand behind it; 68% want to see real people. The platforms have read the same data. Since July, Snapchat no longer recommends wholly AI-generated videos in Spotlight. That is a stricter line than TikTok, YouTube or Meta, which mostly label and deprioritise rather than exclude.</p>\n<p>Meanwhile the advertising middle is quietly saturated. IAB&#39;s 2025 video research found a majority of ad buyers already using generative AI for video creation, 30% of digital video ads built from scratch or enhanced with it, and a forecast of 39% for 2026; its separate July 2026 report puts US digital video ad spend past $80bn. Generated video is not a differentiator in that market. It is a cost floor.</p>\n<p>The Sphere show, by contrast, has sold more than three million tickets and taken over $400m since it opened in August 2025. Nobody bought those tickets to watch AI. They bought them to be inside something at a scale that cannot be reproduced on a phone. Generative models were how a 1939 negative was made big enough to fill it.</p>\n<h2 id=\"what-a-generated-frame-actually-costs-on-a-wall\">What a generated frame actually costs on a wall</h2>\n<p>At BCVR we have watched the same line-item collapse everyone else has. Concepting, previs, texture and plate work that used to consume weeks of a schedule now resolve in an afternoon, and on some projects the raw generation cost has fallen far enough that it no longer appears as its own budget line.</p>\n<p>Almost none of that saving reaches the bottom of the sheet, because it was never where the cost lived. The money on a hybrid physical-digital build goes into warping and blending across projector or LED batches that never quite match; colour-managing generated content that was trained on consumer footage and arrives with the wrong gamma; cutting loop seams that hold up on the four-hundredth pass in front of a queue; provenance paperwork for every generated asset in a client&#39;s brand-safety review; and the on-site week when someone has to look at the actual wall and say it is not right yet. What generative models changed is not the price of the room. It is how many ideas a studio can afford to test before committing to one, and that shows up, when we measure it, as dwell time and throughput rather than as render savings.</p>\n<h2 id=\"buy-the-pipeline-not-the-prompt\">Buy the pipeline, not the prompt</h2>\n<p>The lesson of the Sora sunset is not that generative video is overrated. It is a governance point with a date on it: a capability many teams treated as permanent infrastructure had a discontinuation notice attached, and anyone whose workflow assumed that specific endpoint has one month to move. Model-agnostic pipelines, local copies of what you generated, and provenance records that survive a vendor&#39;s strategic pivot are no longer good hygiene. They are the difference between an asset library and a hostage.</p>\n<p>For studios and commissioners the practical read is simpler still. Generation is now the commodity input; the defensible work is everything downstream of it, in the physical envelope where the audience actually stands. Price the room, not the render.</p>\n<p><em>Scott Marshall is founder of BCVR (<a href=\"https://www.bcvr.io\">https://www.bcvr.io</a>), a London-based immersive production studio working across XR, projection mapping, spatial audio and digital twins for brand and cultural clients.</em></p>\n<h2 id=\"sources\">Sources</h2>\n<ol>\n<li><a href=\"https://help.openai.com/en/articles/20001152-what-to-know-about-the-sora-discontinuation\" target=\"_blank\" rel=\"noopener\">OpenAI Help Center, &quot;What to know about the Sora discontinuation&quot;: app discontinued 26 April 2026; API discontinued 24 September 2026</a></li>\n<li><a href=\"https://www.emarketer.com/content/openai-discontinue-sora-app-disney-exits-partnership\" target=\"_blank\" rel=\"noopener\">eMarketer, &quot;OpenAI to discontinue Sora app as Disney exits partnership&quot; (announced 25 March 2026; App Store #1 to #126; Disney three-year deal, planned $1bn stake, 200+ characters)</a></li>\n<li><a href=\"https://en.wikipedia.org/wiki/Sora_%28text-to-video_model%29\" target=\"_blank\" rel=\"noopener\">Wikipedia, &quot;Sora (text-to-video model)&quot;: release and discontinuation timeline</a></li>\n<li><a href=\"https://www.prnewswire.com/news-releases/higgsfield-raises-400-million-series-b-financing-at-5-4-billion-valuation-with-annualized-revenue-reaching-700-million-302852430.html\" target=\"_blank\" rel=\"noopener\">PR Newswire, &quot;Higgsfield Raises $400 Million Series B Financing at $5.4 Billion Valuation…&quot;, 17 August 2026</a></li>\n<li><a href=\"https://deepmind.google/models/veo/\" target=\"_blank\" rel=\"noopener\">Google DeepMind, Veo model page: 1080p and 4K output, eight-second generations, scene extension</a></li>\n<li><a href=\"https://openrouter.ai/google/veo-3.1\" target=\"_blank\" rel=\"noopener\">OpenRouter, Veo 3.1 listing: $0.40/second with audio, $0.20/second video-only; up to 20 chained clips for 140+ second narratives</a></li>\n<li><a href=\"https://blog.google/innovation-and-ai/infrastructure-and-cloud/google-cloud/sphere-wizard-of-oz/\" target=\"_blank\" rel=\"noopener\">Google, &quot;How Sphere&#39;s new &#39;The Wizard of Oz&#39; experience is coming to life with AI&quot;</a></li>\n<li><a href=\"https://www.thesphere.com/shows/wizard-of-oz-experience/ai\" target=\"_blank\" rel=\"noopener\">Sphere, &quot;The Wizard of Oz at Sphere: AI&quot; (160,000 sq ft, 16K x 16K)</a></li>\n<li><a href=\"https://www.tvbeurope.com/artificial-intelligence/how-google-used-ai-to-reconceptualise-the-wizard-of-oz\" target=\"_blank\" rel=\"noopener\">TVBEurope, &quot;How Google used AI to &#39;reconceptualise&#39; The Wizard of Oz&quot;: Buzz Hayes quote</a></li>\n<li><a href=\"https://investor.sphereentertainmentco.com/press-releases/news-details/2026/The-Wizard-of-Oz-at-Sphere-Surpasses-400-Million-in-Ticket-Sales-With-Over-3-Million-Total-Tickets-Sold/default.aspx\" target=\"_blank\" rel=\"noopener\">Sphere Entertainment Co. investor release, 16 June 2026: 3m+ tickets, $400m+ ticket sales since 28 August 2025</a></li>\n<li><a href=\"https://neon.reviewjournal.com/kats/report-wizard-production-would-cost-sphere-80m-3118777/\" target=\"_blank\" rel=\"noopener\">Las Vegas Review-Journal / Neon, 5 August 2024, reporting an unconfirmed New York Post figure of $80m against an 80-minute plan</a></li>\n<li><a href=\"https://en.wikipedia.org/wiki/The_Wizard_of_Oz_at_Sphere\" target=\"_blank\" rel=\"noopener\">Wikipedia, &quot;The Wizard of Oz at Sphere&quot;: 75-minute runtime, ~1.2 petabytes, $100m budget figure</a></li>\n<li><a href=\"https://www.businesswire.com/news/home/20260121875037/en/83-of-Consumers-Can-Spot-AI-Videos-36-Say-It-Lowers-Brand-Trust-According-to-Animotos-New-Report\" target=\"_blank\" rel=\"noopener\">Business Wire, Animoto 2026 State of Video Report, 21 January 2026 (450+ respondents, surveyed September 2025)</a></li>\n<li><a href=\"https://www.marketingdive.com/news/over-half-of-ad-buyers-are-using-generative-ai-for-video-creation-iab/753048/\" target=\"_blank\" rel=\"noopener\">Marketing Dive, 18 July 2025, reporting IAB&#39;s 2025 Digital Video Ad Spend &amp; Strategy Report (Part Two): 30% of digital video ads, forecast 39% for 2026</a></li>\n<li><a href=\"https://www.iab.com/insights/video-ad-spend-report-2026/\" target=\"_blank\" rel=\"noopener\">IAB, 2026 Digital Video Ad Spend &amp; Strategy Report, 14 July 2026: US digital video ad spend surpasses $80bn in 2026</a></li>\n<li><a href=\"https://mlq.ai/news/snapchat-cuts-fully-ai-generated-videos-from-spotlight-recommendations/\" target=\"_blank\" rel=\"noopener\">MLQ News, &quot;Snapchat cuts fully AI-generated videos from Spotlight recommendations&quot;, July 2026</a></li>\n</ol>\n",
      "image": "https://www.bcvr.io/blog/images/the-clip-is-free/hero-1200x630.jpg",
      "date_published": "2026-08-25T12:00:00.000Z",
      "date_modified": "2026-08-25T12:00:00.000Z",
      "tags": [
        "Industry",
        "AI film",
        "Installations"
      ],
      "authors": [
        {
          "name": "Scott Marshall"
        }
      ]
    },
    {
      "id": "https://www.bcvr.io/blog/the-deployer-in-the-room/",
      "url": "https://www.bcvr.io/blog/the-deployer-in-the-room/",
      "title": "The Deployer in the Room",
      "summary": "Immersive work built its case on generative content and audience sensing. On 2 August, both became somebody's legal responsibility, and that somebody is standing on site.",
      "content_html": "<p>On 12 August, the German non-profit HateAid filed a criminal complaint against Meta, its eyewear partners Ray-Ban and Oakley, and several retailers, arguing that selling glasses capable of covert recording is itself an offence under Germany&#39;s telecommunications and data protection law. The group&#39;s specific objection is elegant: the glasses do carry an LED that flashes while recording, but aftermarket accessories can smother that light without tripping Meta&#39;s anti-tamper safeguard. HateAid wants recording restricted by place, the way dashcams and bodycams already are.</p>\n<p>Ten days earlier, and far more quietly, Article 50 of the EU AI Act came into force. It is the transparency layer of the regulation, and it binds four things: systems that talk to people, systems that generate synthetic media, systems that read emotion or sort people biometrically, and content that depicts real people or places convincingly enough to pass. Fines run to EUR 15 million or 3% of worldwide annual turnover, whichever is larger. The AI Omnibus adopted in late July pushed the high-risk regime out to December 2027, creating a widespread and incorrect impression that everything had slipped. Transparency did not.</p>\n<p>Put those two dates together and a picture emerges that the immersive industry has not really looked at yet. The sensing became trivially powerful, the content became trivially cheap, and the legal duty for both landed on whoever switches the room on.</p>\n<h2 id=\"the-sensing-got-good-before-anyone-asked-it-to\">The sensing got good before anyone asked it to</h2>\n<p>The uncomfortable finding of the year came out of an information security journal in April, not a trade show. Working with roughly 407 participants, researchers identified individual users from eye-tracking data alone at 96.61% accuracy during passive video watching, after less than a minute of viewing. The authors put the figure at around 38 seconds, and note pointedly that this is the length of a standard online advert or a short cutscene. Gaze data has separately been shown to leak age, gender, health conditions, sexual orientation and decision-making patterns. Their conclusion is worth quoting because it is not hedged: the barrier to entry for effective surveillance in immersive environments is &quot;dangerously low, requiring neither sophisticated hardware nor active user cooperation.&quot;</p>\n<p>Meanwhile the cameras multiplied. IDC has display-less smart glasses shipping 2.25 million units in the first quarter of 2026 alone, up 167% year on year, on the way to 13.6 million units and $5.1 billion for the full year. Meta holds 69.2% of that category. Average selling price is $376 now and forecast to fall toward $229 by 2030. In other words, camera-equipped eyewear is on a glide path to the price of a promotional giveaway.</p>\n<p>And the measurement layer arrived from inside the industry rather than outside it. At the Experiential Marketing Summit this March, the trade press deployed a behavioural analytics platform across the exhibit hall, using discreetly installed sensors to log attendee movement, dwell time and engagement, with per-exhibitor performance reports afterwards. That is a reasonable thing to do. It is also, in the Act&#39;s vocabulary, a deployment.</p>\n<h2 id=\"the-duty-follows-the-switch-not-the-invoice\">The duty follows the switch, not the invoice</h2>\n<p>The distinction Article 50 draws is between providers and deployers, and it is the single most important thing for a studio to understand.</p>\n<p>Providers (the people who build the model or the sensing system) must mark synthetic outputs in a machine-readable, detectable format, and must design conversational systems so people know they are talking to software. Deployers carry two separate duties. If an AI system recognises emotions or categorises people biometrically, the deployer must inform the people exposed to it. If content depicting real people or places is AI-generated or manipulated, the deployer discloses it.</p>\n<p>The deployer is not the vendor who sold the sensor bar. It is the party operating the system in the room: the venue, the brand, and very often the production studio that specified, installed and commissioned it. Disclosure has to arrive at the latest at the moment of first exposure, in a form described as clear and distinguishable. Buried text, vague labelling and a line in the terms and conditions are explicitly named as insufficient. That rules out the industry&#39;s default answer, a laminated A4 notice by the entrance that nobody reads in a dark room.</p>\n<p>There is one date still in the diary. Generative systems already on the market get until 2 December 2026 to comply with the marking and detection requirement. Everything else has been live since 2 August.</p>\n<blockquote class=\"pull\"><p>The compliance question is no longer &quot;did we buy a compliant tool?&quot; but &quot;who was standing at the door when it switched on?&quot;</p>\n</blockquote><h2 id=\"what-this-looks-like-on-a-build\">What this looks like on a build</h2>\n<p>At BCVR we specify sensing on a decent share of projects, because clients now ask what the installation did rather than whether it looked good, and dwell time and flow are the only honest answers available. The change since 2 August is not philosophical, it is a line-items change, and it is smaller than people fear.</p>\n<p>Three things move. Signage becomes a design problem rather than an afterthought: notification at the threshold, legible in the light level you have actually built, which for a projection-mapped space is close to none. Second, the capture policy gets written before the kit is chosen. Anonymised movement analytics and biometric categorisation are very different animals under the Act, and a lot of vendor decks blur them. Third, generative content acquires a provenance trail. The cost collapse in AI-generated visual assets is real and it is the main reason room-scale work has become affordable for mid-tier cultural clients, but marked, machine-readable output is now the price of using it inside the EU.</p>\n<p>None of this is expensive: perhaps a day of production coordination and a conversation with the client&#39;s data protection lead. What it requires is that someone owns it, and on most immersive builds nobody currently does. It falls between the AV integrator, the agency and the venue, which is precisely the gap the regulation is written to close.</p>\n<h2 id=\"the-bystander-became-a-stakeholder\">The bystander became a stakeholder</h2>\n<p>The wider shift is that the person who never opted in now has standing.</p>\n<p>The pattern is visible across three unrelated fronts this year: the HateAid complaint in Germany in August; a US class action filed in the Northern District of California in March, alleging that Ray-Ban smart glasses were marketed on privacy claims while subcontracted workers reviewed captured footage; and the security conference DEF CON banning camera glasses outright in July. Legal commentators have identified the common thread: privacy law has historically been framed as disclosure to the purchaser, leaving the bystander&#39;s interests thinly protected. That is the gap now closing, and Article 50&#39;s emotion-recognition clause closes it by addressing the exposed person, not the customer.</p>\n<p>Britain sits outside the AI Act but not outside the direction of travel. The ICO&#39;s March strategy update names automated decision-making, foundation models and facial recognition as priorities and is engaging eleven major model developers directly, with police facial recognition audits already run at forces including West Yorkshire and Greater Manchester.</p>\n<p>The practical read for anyone commissioning or building immersive work: assume that within the next two procurement cycles, a capture-and-disclosure schedule will be a standard annex in venue contracts, the way public liability and rigging certification already are. The studios that have one drafted will win the work. The ones treating this as a legal problem to be discovered during install will lose a fortnight they did not budget.</p>\n<h2 id=\"sources\">Sources</h2>\n<ol>\n<li><a href=\"https://www.cooley.com/news/insight/2026/2026-08-03-eu-ai-act-transparency-obligations-take-effect-2-august-2026\" target=\"_blank\" rel=\"noopener\">Cooley, &quot;EU AI Act: Transparency Obligations Take Effect 2 August 2026,&quot; 3 Aug 2026</a></li>\n<li><a href=\"https://www.goodwinlaw.com/en/insights/publications/2026/08/alerts-technology-dpc-eu-ai-act-transparency-obligations-now-in-force\" target=\"_blank\" rel=\"noopener\">Goodwin, &quot;Not Delayed, Not Deferred: EU AI Act Transparency Obligations Are Now in Force,&quot; Aug 2026</a></li>\n<li><a href=\"https://artificialintelligenceact.eu/transparency-rules-article-50/\" target=\"_blank\" rel=\"noopener\">EU Artificial Intelligence Act, &quot;A Practical Guide to Article 50&quot;</a></li>\n<li><a href=\"https://link.springer.com/article/10.1007/s10207-026-01231-3\" target=\"_blank\" rel=\"noopener\">&quot;Through the looking glass: eye tracking biometrics and the loss of anonymity in extended reality,&quot; Int. J. Information Security, 1 Apr 2026</a></li>\n<li><a href=\"https://www.idc.com/resource-center/blog/smart-glasses-surge-the-xr-market-is-rewriting-its-own-rules/\" target=\"_blank\" rel=\"noopener\">IDC, &quot;Smart Glasses Surge: The XR Market Is Rewriting Its Own Rules&quot;</a></li>\n<li><a href=\"https://www.engadget.com/2235857/german-nonprofit-files-criminal-complaint-over-meta-smart-glasses-privacy/\" target=\"_blank\" rel=\"noopener\">Engadget, &quot;German nonprofit files criminal complaint over Meta smart glasses privacy,&quot; 12 Aug 2026</a></li>\n<li><a href=\"https://fortune.com/2026/07/11/meta-ray-ban-smart-glasses-camera-led-light-privacy-safeguard-super-sensing-ai-prototype-covert-recording-concerns/\" target=\"_blank\" rel=\"noopener\">Fortune, &quot;Meta tightens Ray-Ban smart glasses privacy while it tests &#39;super-sensing&#39; AI prototype,&quot; 11 Jul 2026</a></li>\n<li><a href=\"https://natlawreview.com/article/smart-glasses-and-privacy-wearable-surveillance-and-disclosure-issues\" target=\"_blank\" rel=\"noopener\">National Law Review, &quot;Smart Glasses and Privacy: Wearable Surveillance and Disclosure Issues,&quot; 9 Apr 2026</a></li>\n<li><a href=\"https://www.accessnewswire.com/newsroom/en/computers-technology-and-internet/event-marketer-to-measure-attendee-engagement-with-fastsensor-at-1148775\" target=\"_blank\" rel=\"noopener\">AccessNewswire, &quot;Event Marketer to Measure Attendee Engagement with FastSensor at EMS 2026,&quot; 18 Mar 2026</a></li>\n<li><a href=\"https://ico.org.uk/about-the-ico/our-information/our-strategies-and-plans/artificial-intelligence-and-biometrics-strategy/ai-and-biometrics-strategy-update-march-2026/\" target=\"_blank\" rel=\"noopener\">ICO, &quot;AI and biometrics strategy update: March 2026&quot;</a></li>\n</ol>\n",
      "image": "https://www.bcvr.io/blog/images/the-deployer-in-the-room/hero-1200x630.jpg",
      "date_published": "2026-08-19T12:00:00.000Z",
      "date_modified": "2026-08-19T12:00:00.000Z",
      "tags": [
        "Immersive",
        "XR",
        "Installations",
        "Industry"
      ],
      "authors": [
        {
          "name": "Scott Marshall"
        }
      ]
    },
    {
      "id": "https://www.bcvr.io/blog/the-interface-with-no-screen/",
      "url": "https://www.bcvr.io/blog/the-interface-with-no-screen/",
      "title": "The Interface with No Screen",
      "summary": "The fastest-growing device in extended reality doesn't show you anything. For everyone building immersive work, that changes the brief.",
      "content_html": "<p>Between January and March this year, the world bought 2.25 million pairs of smart glasses with no display in them. That is a 167% increase on the same quarter a year earlier, and it is very nearly the entire category&#39;s volume for all of 2024 compressed into thirteen weeks.</p>\n<p>In the same stretch, Meta&#39;s Reality Labs (the division that has spent a decade arguing that immersive computing is something you look through) reported a second-quarter operating loss of $4.62 billion against $431 million of revenue. That is wider than the $4.03 billion it lost in the first quarter, and it takes cumulative losses since late 2020 to roughly $88 billion.</p>\n<p>Both numbers describe the same industry. Only one of them involves a screen.</p>\n<p>The first genuinely mass-market device in extended reality is not a display you wear. It is a microphone and a pair of open-ear speakers with a language model sitting behind them. It has quietly outrun an immersive sector whose entire vocabulary, toolchain and cost base is visual: an interface that sector barely staffs for.</p>\n<h2 id=\"the-quarter-that-outsold-a-year\">The quarter that outsold a year</h2>\n<p>IDC&#39;s mid-June figures reward a close read. Display-less smart glasses are forecast at 13.6 million units and $5.1 billion of revenue for 2026, rising to 27.3 million units by 2030. Optical see-through display glasses (the transparent overlay the industry has promised for fifteen years) come in at 3 million units this year. Mixed-reality headsets: 3.2 million units, $2.4 billion.</p>\n<p>Put plainly, roughly two out of every three devices shipping under the XR banner in 2026 have nothing to show you. Meta holds 69.2% of that display-less segment. Average selling price sits at $376 and is forecast to fall to about $229 by 2030, which is the price curve of a consumer accessory rather than a computing platform.</p>\n<p>The wider market has the same shape. IDC counted 14.5 million XR devices shipped in 2025, up 41.6%. It is a healthy headline that conceals a 42.8% collapse in MR and VR headsets set against 211.2% growth across everything else.</p>\n<h2 id=\"every-platform-is-converging-on-the-ear\">Every platform is converging on the ear</h2>\n<p>On 23 June, Meta launched its first self-branded eyewear with EssilorLuxottica: Meta Glasses, from $299, twenty-six styles at launch, open-ear speakers, a multi-microphone array with wind-noise reduction, and its Muse Spark model aboard from day one. There is no display. Pedestrian navigation, the company notes, is &quot;coming soon for displayless glasses.&quot;</p>\n<p>Apple, according to reporting from Bloomberg and Reuters relayed through the analyst chain, shelved the cheaper Vision Air headset to prioritise lightweight AI glasses that explicitly do not have transparent AR displays. What they do have: cameras, audio, live translation, environmental sensing. Announcement this year, launch expected 2027; glasses with colour displays are not on the roadmap until 2028.</p>\n<p>Samsung told its July earnings call it would ship &quot;Intelligent eyewear&quot; later this year, developed with Google on Android XR. Note the noun. Not glasses, not headset, not XR: a category name chosen to avoid every word the industry has been using, in the same quarter the company posted the first loss its mobile division has ever recorded.</p>\n<p>Three companies, three strategies, one convergent bet: the thing people will actually wear on their face all day is a listening device.</p>\n<blockquote class=\"pull\"><p>&quot;The medium arrived on schedule. It just came through the ear, and most of us were watching the eye.&quot;</p>\n</blockquote><h2 id=\"the-venues-knew-first\">The venues knew first</h2>\n<p>Anyone who has commissioned a large physical installation will find none of this surprising. The high end of immersive has been audio-led for years, quietly, while the coverage went to the pixels.</p>\n<p>The Sphere in Las Vegas runs 167,000 individually amplified speaker drivers behind its interior surfaces, using Holoplot&#39;s beamforming matrix arrays. At Lightroom in King&#39;s Cross, David Bowie: You&#39;re Not Alone ran from 22 April to 28 June on two Holoplot X1 Matrix Arrays (four Modul 96s and four Modul 80-Ss apiece) recessed into the wall cavities, with the show&#39;s tracks re-authored for the room by sound designer Gareth Fry. Festival-scale object-based rigs now run past 170 loudspeakers in a single L-ISA deployment.</p>\n<p>Nobody has ever sold a ticket to a wavefield synthesis array. They sell the feeling it produces. Audiences almost always describe that feeling in visual language, saying they felt inside something, because English is short of vocabulary for what accurate spatialisation does to a room.</p>\n<h2 id=\"what-it-costs-from-the-floor\">What it costs, from the floor</h2>\n<p>Three things look different from inside a production studio than they do from a market forecast.</p>\n<p>The first is that in hybrid physical-digital work, sound is the cheapest immersion per pound spent. A room&#39;s audio layer can be re-authored in days; its structure, surfaces and projection content cannot. When a client wants a second life out of an installation, the audio is where the change is affordable.</p>\n<p>The second is that generative AI has collapsed the cost of visual content specifically, and almost nothing else. That asymmetry matters more than the raw saving. If images, textures and 3D assets are trending toward free, the visual layer is the part that commoditises first; the differentiator migrates to what remains expensive and hard to fake: how a space sounds, and how precisely it responds to the person standing in it.</p>\n<p>The third is measurement, where the sector remains weakest. Industry budget guidance puts a single interactive station at roughly $20,000 to $50,000 and a room-scale multi-station zone at $40,000 to $150,000 and up, with vendors claiming well-designed interactives lift dwell time two to five times over static display. Treat the multiplier as a supplier&#39;s own figure rather than independent research, but note what it measures. Dwell time is precisely the metric a considered spatial audio layer moves, without touching the build, the structure or the schedule.</p>\n<p>That is the practical opening. A brand or a museum that cannot justify headset logistics for a thousand visitors a day can justify a soundfield, and increasingly can justify content that follows the visitor out of the building through eyewear they already own: no app install, no hardware to hand out, no queue for a headset wipe.</p>\n<h2 id=\"what-to-do-with-it\">What to do with it</h2>\n<p>If the installed base arriving on people&#39;s faces this year is audio-first, three things follow for anyone commissioning or building immersive work. Specify object-based audio in the treatment rather than the snagging list, because retrofitting spatialisation costs multiples of designing for it. Bring a sound designer in at concept stage, on the same call as the creative director, not at the mix. And when scoping the take-home layer of a physical experience, assume the device in the audience&#39;s pocket or on their nose can hear and speak, and cannot show. That is a constraint, and a considerably cheaper one than the alternative.</p>\n<p>The industry spent a decade waiting for the display to get good enough. It may turn out the display was the expensive way to do it.</p>\n<p>Scott is the founder of BCVR (<a href=\"https://www.bcvr.io\">https://www.bcvr.io</a>), a London-based immersive experiential production studio working across XR, projection mapping, spatial audio and digital twins for brands and cultural institutions.</p>\n<h2 id=\"sources\">Sources</h2>\n<ol>\n<li><a href=\"https://www.idc.com/resource-center/blog/smart-glasses-surge-the-xr-market-is-rewriting-its-own-rules/\" target=\"_blank\" rel=\"noopener\">IDC, &quot;Smart Glasses Market 2026: XR Is Rewriting the Rules,&quot; 15 Jun 2026</a></li>\n<li><a href=\"https://vr.org/articles/meta-reality-labs-q2-2026-earnings-loss-widens-88-billion\" target=\"_blank\" rel=\"noopener\">VR.org, &quot;Reality Labs Lost $4.62 Billion in Q2,&quot; 31 Jul 2026</a></li>\n<li><a href=\"https://my.idc.com/getdoc.jsp?containerId=prUS54033425\" target=\"_blank\" rel=\"noopener\">IDC, &quot;Global XR Shipments Rebound Behind Glasses-First Momentum,&quot; 11 Dec 2025</a></li>\n<li><a href=\"https://about.fb.com/news/2026/06/meta-essilorluxottica-partner-launch-meta-glasses/\" target=\"_blank\" rel=\"noopener\">Meta Newsroom, &quot;We&#39;re Partnering With EssilorLuxottica to Launch Meta Glasses,&quot; 23 Jun 2026</a></li>\n<li><a href=\"https://9to5google.com/2026/07/30/samsung-teases-android-xr-glasses-galaxy-tab-s12-s26-fe-and-more/\" target=\"_blank\" rel=\"noopener\">9to5Google, &quot;Samsung teases Android XR glasses…,&quot; 30 Jul 2026</a></li>\n<li><a href=\"https://apple.gadgethacks.com/news/apple-scraps-vision-air-for-ai-smart-glasses-in-2026/\" target=\"_blank\" rel=\"noopener\">Gadget Hacks, &quot;Apple Scraps Vision Air for AI Smart Glasses in 2026,&quot; Oct 2025 (reporting Bloomberg/Reuters)</a></li>\n<li><a href=\"https://www.ticketfairy.com/blog/immersive-audio-in-2026-how-spatial-sound-tech-is-elevating-live-events\" target=\"_blank\" rel=\"noopener\">Ticket Fairy, &quot;Immersive Audio in 2026,&quot; 21 Jan 2026</a></li>\n<li><a href=\"https://www.installation-international.com/events/visitor-attractions/holoplot-spatial-audio-installed-at-lightroom-for-bowie-exhibition\" target=\"_blank\" rel=\"noopener\">Installation International, &quot;HOLOPLOT spatial audio installed at Lightroom for Bowie exhibition,&quot; 19 Feb 2026</a></li>\n<li><a href=\"https://www.utsubo.com/blog/interactive-museum-installations-benefits-guide\" target=\"_blank\" rel=\"noopener\">Utsubo, &quot;Interactive Museum Installations: ROI, Costs &amp; Real Examples (2026)&quot;</a></li>\n</ol>\n",
      "image": "https://www.bcvr.io/blog/images/the-interface-with-no-screen/hero-1200x630.jpg",
      "date_published": "2026-08-17T12:00:00.000Z",
      "date_modified": "2026-08-17T12:00:00.000Z",
      "tags": [
        "XR",
        "Spatial sound",
        "Industry"
      ],
      "authors": [
        {
          "name": "Scott Marshall"
        }
      ]
    }
  ]
}
