--- title: The BEST local AI video generator is here! | daily.dev description: MiniMax H3 is a new open-source multimodal video generation model that can run locally via ComfyUI. It supports text-to-video, image-to-video, and... canonical: https://daily.dev/posts/the-best-local-ai-video-generator-is-here--a2oalliis twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: The BEST local AI video generator is here! | daily.dev og:description: MiniMax H3 is a new open-source multimodal video generation model that can run locally via ComfyUI. It supports text-to-video, image-to-video, and... og:url: https://daily.dev/posts/the-best-local-ai-video-generator-is-here--a2oalliis og:image: https://api.daily.dev/og/posts/A2OAlLIIs.png og:image:alt: The BEST local AI video generator is here! 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. # The BEST local AI video generator is here! **[YouTube](https://daily.dev/sources/youtube)** · 23 min read · 0 upvotes · 0 comments ## Summary MiniMax H3 is a new open-source multimodal video generation model that can run locally via ComfyUI. It supports text-to-video, image-to-video, and reference-to-video workflows, with built-in audio generation and strong character consistency. The full model is 66 GB but pruned versions (21 GB) can run on as little as 12 GB VRAM, with a third-party platform enabling 5 GB VRAM usage at 480p. The tutorial covers downloading models (diffusion model, Qwen-3-VL-32B text encoder, audio/video VAEs), setting up ComfyUI workflows, and speed optimizations using SageAttention, easy cache nodes, and ComfyUI Spectrum. Licensing is permissive for commercial use under $20M revenue, but users in the EU, UK, Korea, and US must apply for a separate license. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### What are the VRAM requirements to run MiniMax H3 locally? MiniMax H3 can run on as little as 5 GB VRAM using the 1toGP platform at 480p resolution. With ComfyUI, the pruned FP8 or INT8 models (21 GB) have been reported to work on 12 GB VRAM. The full model is 66 GB, an INT8 version is 34 GB, and pruned versions are 21 GB, with quality decreasing as compression increases. _Developers running local AI video generation track hardware compatibility updates and new model releases on daily.dev._ ### What models do I need to download to run MiniMax H3 in ComfyUI? Running MiniMax H3 in ComfyUI requires three components: a diffusion model (FL2VA for text/image-to-video or Ref2VA for reference-to-video, available in full 66 GB, INT8 34 GB, or pruned FP8/INT8 21 GB), the Qwen-3-VL-32B text encoder (also available in compressed versions, can run on RAM rather than VRAM), and both an audio VAE and a video VAE placed in the ComfyUI models/VAE folder. _Those setting up local AI pipelines find workflow-specific model requirements like these covered on daily.dev._ ### How can I speed up MiniMax H3 video generation in ComfyUI? Three methods can speed up MiniMax H3 generation in ComfyUI. First, installing SageAttention and adding a 'patch sage attention' node after the load diffusion model node yields 20–30% faster generation. Second, adding an 'easy cache' node between diffusion model nodes adds another 5–10%. Third, installing ComfyUI Spectrum, which is specifically designed for MiniMax H3, provides additional speedup. All three can be stacked together. _Optimizing local AI video generation workflows is the kind of hands-on topic developers follow on daily.dev._ --- Tags: [#open-source](https://daily.dev/tags/open-source), [#local-ai](https://daily.dev/tags/local-ai), [#ai-video-generation](https://daily.dev/tags/ai-video-generation) [View this post on daily.dev](https://daily.dev/posts/the-best-local-ai-video-generator-is-here--a2oalliis) ```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":"The BEST local AI video generator is here!","url":"https://daily.dev/posts/the-best-local-ai-video-generator-is-here--a2oalliis","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/the-best-local-ai-video-generator-is-here--a2oalliis"},"datePublished":"2026-08-10T21:18:53.314Z","dateModified":"2026-08-14T17:59:19.385Z","description":"MiniMax H3 is a new open-source multimodal video generation model that can run locally via ComfyUI. It supports text-to-video, image-to-video, and...","image":"https://i.ytimg.com/vi/8HutJ9W5pTg/sddefault.jpg","thumbnailUrl":"https://i.ytimg.com/vi/8HutJ9W5pTg/sddefault.jpg","isAccessibleForFree":true,"articleSection":"YouTube","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":"YouTube","logo":"https://media.daily.dev/image/upload/s--W5zMumpP--/f_auto/v1725350181/logos/youtube","url":"https://daily.dev/sources/youtube"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/the-best-local-ai-video-generator-is-here--a2oalliis","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"open-source,local-ai,ai-video-generation","timeRequired":"PT23M","video":{"@type":"VideoObject","name":"The BEST local AI video generator is here!","description":"MiniMax H3 is a new open-source multimodal video generation model that can run locally via ComfyUI. It supports text-to-video, image-to-video, and...","thumbnailUrl":"https://i.ytimg.com/vi/8HutJ9W5pTg/sddefault.jpg","uploadDate":"2026-08-10T21:18:53.314Z","duration":"PT23M","url":"https://api.daily.dev/r/A2OAlLIIs","embedUrl":"https://www.youtube.com/embed/8HutJ9W5pTg"}} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"YouTube","item":"https://daily.dev/sources/youtube"},{"@type":"ListItem","position":3,"name":"The BEST local AI video generator is here!"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/the-best-local-ai-video-generator-is-here--a2oalliis#faq","mainEntity":[{"@type":"Question","name":"What are the VRAM requirements to run MiniMax H3 locally?","acceptedAnswer":{"@type":"Answer","text":"MiniMax H3 can run on as little as 5 GB VRAM using the 1toGP platform at 480p resolution. With ComfyUI, the pruned FP8 or INT8 models (21 GB) have been reported to work on 12 GB VRAM. The full model is 66 GB, an INT8 version is 34 GB, and pruned versions are 21 GB, with quality decreasing as compression increases. Developers running local AI video generation track hardware compatibility updates and new model releases on daily.dev."}},{"@type":"Question","name":"What models do I need to download to run MiniMax H3 in ComfyUI?","acceptedAnswer":{"@type":"Answer","text":"Running MiniMax H3 in ComfyUI requires three components: a diffusion model (FL2VA for text/image-to-video or Ref2VA for reference-to-video, available in full 66 GB, INT8 34 GB, or pruned FP8/INT8 21 GB), the Qwen-3-VL-32B text encoder (also available in compressed versions, can run on RAM rather than VRAM), and both an audio VAE and a video VAE placed in the ComfyUI models/VAE folder. Those setting up local AI pipelines find workflow-specific model requirements like these covered on daily.dev."}},{"@type":"Question","name":"How can I speed up MiniMax H3 video generation in ComfyUI?","acceptedAnswer":{"@type":"Answer","text":"Three methods can speed up MiniMax H3 generation in ComfyUI. First, installing SageAttention and adding a 'patch sage attention' node after the load diffusion model node yields 20–30% faster generation. Second, adding an 'easy cache' node between diffusion model nodes adds another 5–10%. Third, installing ComfyUI Spectrum, which is specifically designed for MiniMax H3, provides additional speedup. All three can be stacked together. Optimizing local AI video generation workflows is the kind of hands-on topic developers follow on daily.dev."}}]} ```