--- title: Google DeepMind announces Genie 4, a world model with... description: Google DeepMind announced Genie 4, a world model that generates interactive environments in real time with synchronized audio. The new version adds ambient... canonical: https://daily.dev/posts/google-deepmind-announces-genie-4-a-world-model-with-synchronized-audio-jf2v69ddy twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Google DeepMind announces Genie 4, a world model with synchronized audio | daily.dev og:description: Google DeepMind announced Genie 4, a world model that generates interactive environments in real time with synchronized audio. The new version adds ambient... og:url: https://daily.dev/posts/google-deepmind-announces-genie-4-a-world-model-with-synchronized-audio-jf2v69ddy og:image: https://api.daily.dev/og/posts/JF2v69ddY.png og:image:alt: Google DeepMind announces Genie 4, a world model with synchronized audio og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Google DeepMind announces Genie 4, a world model with synchronized audio **[Collections](https://daily.dev/sources/collections)** · 3 min read · 3 upvotes · 0 comments ## Summary Google DeepMind announced Genie 4, a world model that generates interactive environments in real time with synchronized audio. The new version adds ambient sound and character dialogue built on the Omni Flash foundation. A separate model, Gemini 3.8 Flash, interprets the scene and drives character responses as it unfolds. The team says frame rate typically runs 24 to 30 FPS rather than 60 because of latency constraints. Outputs carry SynthID watermarks, and access will roll out first to Trusted Testers, then Project Genie and the API. DeepMind names robotics training data and education as target uses. ## Content Google DeepMind has announced Genie 4, a world model that generates interactive, photorealistic environments in real time. It is the first Genie with native audio: ambient sound, music and dialogue from characters who react to the user and to events in the scene. It opened October 27 as a research preview for trusted testers. ## What it does You move through a generated world with keyboard controls, and you can rewrite it while it runs by typing. Prompts can change the weather, move the camera, or add objects and people. Gemini fits each edit to the existing scene so it stays coherent. Scenes now hold together for up to two minutes, twice Genie 3's limit. DeepMind also cites better physics, including collisions, and more human-like characters. The team says output typically runs at 24 to 30 FPS rather than 60, a latency tradeoff. ## How the pieces fit Several Gemini models drive it. Gemini Omni, built on the Omni Flash foundation, handles photorealism and how worlds look and behave. Gemini 3.8 Flash works as an in-loop director: it splits goals into steps, watches incoming frames and adapts the plan. In a robotics experiment, Gemini 4 Argon used that loop to plan a game of checkers for Apptronik's Apollo humanoid, with Genie 4 rendering the result. DeepMind pitches the model as a place for agents and robots to rehearse multi-step tasks, and as a source of training data and educational tools. ## Early hands-on impressions @emollick typed "Neogothic brutalist city inhabited by robots and horses" and got a world in about a minute. He then added dragons and a friendly otter NPC, all in real time. In a Victorian high tea scene, he had a mouse grow wings and fly to a teacup. His verdict: glitchy, but impossibly advanced compared to a couple of years ago. @petergyang built five worlds, including a kaiju attacking San Francisco, a Tokyo ramen stall and a dragon ride to storm a castle. He was more impressed by the live audio and interactivity than the graphics. He thinks world models are in their GPT-3.5 era. In his view the gameplay isn't there yet, since games depend on core loops and progression, but a director could someday generate a whole effects shot from a single frame. ## Limits DeepMind says object permanence and visual detail can drift in longer sessions, interactions last only a few minutes, and real-world places aren't simulated accurately. Accuracy of audio and in-world text isn't solved either. Google hasn't published benchmarks on consistency across repeated runs. ## Access Trusted testers come first. Google AI Ultra subscribers will get it through Project Genie, and early partners through the Gemini API and AI Studio. Ultra starts at $100 a month. No broader release date or API pricing has been announced. All Genie 4 video and audio carries Google's SynthID watermark. ## Questions this post answers ### What new feature does Google DeepMind's Genie 4 world model add compared to earlier versions? Genie 4 adds synchronized audio, including ambient sound and character dialogue, built on the Omni Flash foundation. A second model, Gemini 3.8 Flash, runs alongside the diffusion model to interpret the scene and drive how characters respond as it unfolds. Frame rate runs 24 to 30 FPS rather than 60 due to latency constraints, and outputs carry SynthID watermarks. _Anyone tracking real-time generative world models can follow rollout details like this on daily.dev._ ### Who can access Google DeepMind's Genie 4 world model and for what purposes? Access opens gradually, starting with Trusted Testers before moving to Project Genie and the API. DeepMind points to robotics training data and education as target uses for the real-time interactive environments the model generates. _Developers evaluating new AI tooling for robotics or edtech can track access rollouts like this on daily.dev._ ## Community take How the wider developer community reacted, aggregated from 1 discussion and 173 comments across hackernews (as of 2026-10-29). **TL;DR:** Reaction is largely skeptical, with heavy debate about whether world models will displace game studios (and enable new DRM/streaming lock-in), frustration over Google's closed access and unpolished announcement, and pointed critiques that demo footage still shows glitchy physics and weak object permanence. **Sentiment:** 15% positive · 30% mixed · 55% skeptical **The case for** - Some see real value for robotics/autonomous vehicle simulation training rather than consumer gaming. - A few commenters find the generated prompts and videos genuinely exciting as a glimpse of future capability, cherry-picked or not. - It's viewed as a useful mental-model analog for how humans and robots predict and adapt to their environment. **The pushback** - Many argue closed, inaccessible, quota-limited models with no technical details are hard to get excited about. - Demo clips were criticized for visible physics/continuity failures (disappearing objects, non-toppling statue, no collision on wings). - Several worry this technology primarily enables new forms of DRM/lock-in or cheap ad/content slop rather than better games. - Comparisons to Stadia suggest skepticism that cloud-streamed, compute-heavy interactive experiences are economically or technically viable. - Some mock the announcement's own writing/copy quality (e.g. 'audio-like music') as sloppy or AI-generated without proofreading. **By community** - hackernews (heated): Discussion splits into contentious sub-threads debating business models, access/openness, demo quality, and philosophical tangents, with overall tone leaning skeptical. **Hottest debate:** Whether AI world models will replace traditional game development/ownership (enabling new DRM via cloud streaming) or are fundamentally unworkable the way Stadia was. **Open questions** - Who is the actual target audience for these generated experiences—gamers, filmmakers, or robotics researchers? - Can cloud-rendered world models ever be cost-effective enough to stream to consumers at scale? - Will access and technical details ever be opened up, or will this remain limited to trusted testers and paying partners? **Highlights** > Looks cool, but it's very hard to get me excited about yet another proprietary model for which absolutely zero technical details are released (not even the number of parameters), which is safetymaxxed to the limit, and whose usage quota will (most likely) be so limited that it'll be effectively useless for anyone who doesn't work at Google or doesn't have infinite money. > — [kouteiheika on hackernews · 4 comments](https://news.ycombinator.com/item?id=50261912) > Yawn - another closed world model. Only open weights matter. We should distill everything else we can into open weights. These are exciting for robotics and industrial perhaps, but they have no consumer use case until the lid is blown off and consumers can do whatever they want with them. So much societal progress is behind held up because the leading labs want to keep a stranglehold on the tech. Imagine if every PhD or hacker had access to the raw weights. > — [echelon on hackernews · 1 comments](https://news.ycombinator.com/item?id=50262272) > I don't know if I'm missing the point, but Genie 4 almost looks worse than Genie 3 in terms of player control? It doesn't even seem to respond to inputs, it just looks like they took some AI generated videos and put the key widgets on top. > — [sodacanner on hackernews · 2 comments](https://news.ycombinator.com/item?id=50261519) > That's the thing about machine learning over time, and the big hole in the whole world model idea. Time and time again we make simulations and train models and we think that will work. It never does. And why not? Because the first model that can actually learn to handle the simulation not only can do the simulation internally too, but it can do it better than out simulation. So we get to a point and people keep making complicated simulations of one kind or another ... and then someone who's never used the simulation, and doesn't need it for model training gets by far the highest score on every benchmark and everyone agrees the simulation was kind of beside the point. > — [spwa4 on hackernews](https://news.ycombinator.com/item?id=50262307) > A big chunk of the right side of castle disappeared, including catapult, when camera turned back while flying on dragon. See 1:25 to 1:29. I was not expecting this to happen in a "world model". > — [smusamashah on hackernews · 2 comments](https://news.ycombinator.com/item?id=50261864) **Source threads** - [hackernews](https://news.ycombinator.com/item?id=50261319) · 284 points · 173 comments ## Similar posts on daily.dev - [Google’s Genie 4 world model adds audio, more realistic worlds, and real-time steering](https://daily.dev/posts/google-s-genie-4-world-model-adds-audio-more-realistic-worlds-and-real-time-steering-byu7bdijt) · TechCrunch · 0 upvotes · 0 comments - [Google's Genie 3 world models start to break down after around a minute](https://daily.dev/posts/google-s-genie-3-world-models-start-to-break-down-after-around-a-minute-8opcxsmvb) · GamesIndustry.biz · 2 upvotes · 0 comments - [Google DeepMind connects Street View to Project Genie world model](https://daily.dev/posts/google-deepmind-connects-street-view-to-project-genie-world-model-uak71o9wk) · The Next Web · 1 upvotes · 0 comments --- Tags: [#genai](https://daily.dev/tags/genai), [#robotics](https://daily.dev/tags/robotics), [#google-deepmind](https://daily.dev/tags/google-deepmind), [#world-models](https://daily.dev/tags/world-models) [View this post on daily.dev](https://daily.dev/posts/google-deepmind-announces-genie-4-a-world-model-with-synchronized-audio-jf2v69ddy) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Google DeepMind announces Genie 4, a world model with synchronized audio","url":"https://daily.dev/posts/google-deepmind-announces-genie-4-a-world-model-with-synchronized-audio-jf2v69ddy","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/google-deepmind-announces-genie-4-a-world-model-with-synchronized-audio-jf2v69ddy"},"datePublished":"2026-10-27T16:23:36.440Z","dateModified":"2026-10-29T07:42:04.203Z","description":"Google DeepMind announced Genie 4, a world model that generates interactive environments in real time with synchronized audio. The new version adds ambient...","image":"https://i.ytimg.com/vi/LOP8ul3KqQI/hqdefault.jpg","thumbnailUrl":"https://i.ytimg.com/vi/LOP8ul3KqQI/hqdefault.jpg","isAccessibleForFree":true,"articleSection":"Collections","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Collections","logo":"https://media.daily.dev/image/upload/s--fk_6ycEi--/f_auto,q_auto/v1780996001/logos/collections?_a=BAMAMiWQ0","url":"https://daily.dev/sources/collections"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/google-deepmind-announces-genie-4-a-world-model-with-synchronized-audio-jf2v69ddy","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":3},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"genai,robotics,google-deepmind,world-models","timeRequired":"PT3M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Collections","item":"https://daily.dev/sources/collections"},{"@type":"ListItem","position":3,"name":"Google DeepMind announces Genie 4, a world model with synchronized audio"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/google-deepmind-announces-genie-4-a-world-model-with-synchronized-audio-jf2v69ddy#faq","mainEntity":[{"@type":"Question","name":"What new feature does Google DeepMind's Genie 4 world model add compared to earlier versions?","acceptedAnswer":{"@type":"Answer","text":"Genie 4 adds synchronized audio, including ambient sound and character dialogue, built on the Omni Flash foundation. A second model, Gemini 3.8 Flash, runs alongside the diffusion model to interpret the scene and drive how characters respond as it unfolds. Frame rate runs 24 to 30 FPS rather than 60 due to latency constraints, and outputs carry SynthID watermarks. Anyone tracking real-time generative world models can follow rollout details like this on daily.dev."}},{"@type":"Question","name":"Who can access Google DeepMind's Genie 4 world model and for what purposes?","acceptedAnswer":{"@type":"Answer","text":"Access opens gradually, starting with Trusted Testers before moving to Project Genie and the API. DeepMind points to robotics training data and education as target uses for the real-time interactive environments the model generates. Developers evaluating new AI tooling for robotics or edtech can track access rollouts like this on daily.dev."}}]} ```