--- title: AI Breakthroughs Everyone Missed | daily.dev description: A roundup covers under-the-radar AI releases skipped in bigger weekly news videos. Fish Audio's Drama 3 is a 36-billion-parameter text-to-speech model that... canonical: https://daily.dev/posts/ai-breakthroughs-everyone-missed-p1gyredbf twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: AI Breakthroughs Everyone Missed | daily.dev og:description: A roundup covers under-the-radar AI releases skipped in bigger weekly news videos. Fish Audio's Drama 3 is a 36-billion-parameter text-to-speech model that... og:url: https://daily.dev/posts/ai-breakthroughs-everyone-missed-p1gyredbf og:image: https://api.daily.dev/og/posts/p1GYRedbF.png og:image:alt: AI Breakthroughs Everyone Missed 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. # AI Breakthroughs Everyone Missed **[Matt Wolfe](https://daily.dev/sources/mreflow)** · 25 min read · 0 upvotes · 0 comments ## Summary A roundup covers under-the-radar AI releases skipped in bigger weekly news videos. Fish Audio's Drama 3 is a 36-billion-parameter text-to-speech model that takes plain-English acting directions instead of emotion presets, and needs roughly 46GB VRAM (32GB quantized). ByteDance's Omnihuman 2.0 animates a single photo into a 5-minute talking or singing video at 720p for 7 cents per second. Tencent's Worldcrafter and Skyworks' Matrix Game 4.0 tackle world-model memory loss with camera-position lookup and a separate game engine respectively, while Nvidia's Stream Render H3 repaints a flat color-coded map in real time using data-center GPUs. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### How much VRAM does Fish Audio's Drama 3 text-to-speech model need to run? Fish Audio's Drama 3 is a 36-billion-parameter open-weight text-to-speech model that typically needs about 46GB of VRAM to run. A community-made 4-bit quantized version can run on roughly 32GB, but neither fits on most laptops. The API costs about $10 per million characters at launch, rising to $50 after November 12th. _Developers evaluating TTS models for production can track hardware and pricing shifts like this on daily.dev._ ### What's new in ByteDance's Omnihuman 2.0 talking avatar model compared to the previous version? Omnihuman 2.0 generates continuous talking or singing videos from a single image and audio track for up to five minutes, up from just 60 seconds in the prior version. It renders at 720p and supports text prompts for camera moves. Pricing dropped from 12 cents per second to 7 cents per second, so a full 5-minute video costs about 21 dollars. _Anyone comparing avatar video tools can keep tabs on cost and length limits like these via daily.dev._ ### How does Skyworks Matrix Game 4.0 solve the memory problem in AI world models? Matrix Game 4.0 pairs a traditional game engine, which tracks positions and rules, with a video model that only repaints the engine's blocky low-poly frames to look finished. Because the engine keeps the actual state, objects don't vanish when off-screen. Each scene uses its own roughly 5-billion-parameter model, runs at 480x864, and ships under Apache 2.0 with weights and inference code available. _Developers exploring AI-driven game rendering can follow architecture trade-offs like this on daily.dev._ ## Similar posts on daily.dev - [Voice AI in vogue: Nvidia releases new open transcription model](https://daily.dev/posts/voice-ai-in-vogue-nvidia-releases-new-open-transcription-model-dotdulayg) · TechCrunch · 0 upvotes · 0 comments --- Tags: [#open-source](https://daily.dev/tags/open-source), [#nvidia](https://daily.dev/tags/nvidia), [#video-generation](https://daily.dev/tags/video-generation), [#text-to-speech](https://daily.dev/tags/text-to-speech), [#world-models](https://daily.dev/tags/world-models) [View this post on daily.dev](https://daily.dev/posts/ai-breakthroughs-everyone-missed-p1gyredbf) ```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":"AI Breakthroughs Everyone Missed","url":"https://daily.dev/posts/ai-breakthroughs-everyone-missed-p1gyredbf","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/ai-breakthroughs-everyone-missed-p1gyredbf"},"datePublished":"2026-10-21T17:54:17.473Z","dateModified":"2026-10-21T17:59:51.783Z","description":"A roundup covers under-the-radar AI releases skipped in bigger weekly news videos. Fish Audio's Drama 3 is a 36-billion-parameter text-to-speech model that...","image":"https://i.ytimg.com/vi/umGZrkyTRtw/sddefault.jpg","thumbnailUrl":"https://i.ytimg.com/vi/umGZrkyTRtw/sddefault.jpg","isAccessibleForFree":true,"articleSection":"Matt Wolfe","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":"Matt Wolfe","logo":"https://media.daily.dev/image/upload/s--JWE7fI4Q--/f_auto/v1716024560/logos/mreflow","url":"https://daily.dev/sources/mreflow"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/ai-breakthroughs-everyone-missed-p1gyredbf","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"open-source,nvidia,video-generation,text-to-speech,world-models","timeRequired":"PT25M","video":{"@type":"VideoObject","name":"AI Breakthroughs Everyone Missed","description":"A roundup covers under-the-radar AI releases skipped in bigger weekly news videos. Fish Audio's Drama 3 is a 36-billion-parameter text-to-speech model that...","thumbnailUrl":"https://i.ytimg.com/vi/umGZrkyTRtw/sddefault.jpg","uploadDate":"2026-10-21T17:54:17.473Z","duration":"PT25M","url":"https://api.daily.dev/r/p1GYRedbF","embedUrl":"https://www.youtube.com/embed/umGZrkyTRtw"}} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Matt Wolfe","item":"https://daily.dev/sources/mreflow"},{"@type":"ListItem","position":3,"name":"AI Breakthroughs Everyone Missed"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/ai-breakthroughs-everyone-missed-p1gyredbf#faq","mainEntity":[{"@type":"Question","name":"How much VRAM does Fish Audio's Drama 3 text-to-speech model need to run?","acceptedAnswer":{"@type":"Answer","text":"Fish Audio's Drama 3 is a 36-billion-parameter open-weight text-to-speech model that typically needs about 46GB of VRAM to run. A community-made 4-bit quantized version can run on roughly 32GB, but neither fits on most laptops. The API costs about $10 per million characters at launch, rising to $50 after November 12th. Developers evaluating TTS models for production can track hardware and pricing shifts like this on daily.dev."}},{"@type":"Question","name":"What's new in ByteDance's Omnihuman 2.0 talking avatar model compared to the previous version?","acceptedAnswer":{"@type":"Answer","text":"Omnihuman 2.0 generates continuous talking or singing videos from a single image and audio track for up to five minutes, up from just 60 seconds in the prior version. It renders at 720p and supports text prompts for camera moves. Pricing dropped from 12 cents per second to 7 cents per second, so a full 5-minute video costs about 21 dollars. Anyone comparing avatar video tools can keep tabs on cost and length limits like these via daily.dev."}},{"@type":"Question","name":"How does Skyworks Matrix Game 4.0 solve the memory problem in AI world models?","acceptedAnswer":{"@type":"Answer","text":"Matrix Game 4.0 pairs a traditional game engine, which tracks positions and rules, with a video model that only repaints the engine's blocky low-poly frames to look finished. Because the engine keeps the actual state, objects don't vanish when off-screen. Each scene uses its own roughly 5-billion-parameter model, runs at 480x864, and ships under Apache 2.0 with weights and inference code available. Developers exploring AI-driven game rendering can follow architecture trade-offs like this on daily.dev."}}]} ```