{"id":12744,"date":"2026-11-17T11:01:37","date_gmt":"2026-11-17T02:01:37","guid":{"rendered":"https:\/\/localmodelwatch.tsuchitsuchi.com\/?p=12744"},"modified":"2026-11-17T16:11:44","modified_gmt":"2026-11-17T07:11:44","slug":"fasth3-omni-comfy-released","status":"publish","type":"post","link":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/11\/17\/fasth3-omni-comfy-released\/","title":{"rendered":"FastH3-Omni-Comfy: 80GB+ VRAM, File List"},"content":{"rendered":"<p><!-- lmw:facts --><\/p>\n<h2>At a Glance<\/h2>\n<div class=\"lmw-table-scroll\" tabindex=\"0\" style=\"overflow-x:auto;-webkit-overflow-scrolling:touch;max-width:100%;\">\n<table style=\"width:max-content;min-width:100%;border-collapse:collapse;\">\n<thead>\n<tr>\n<th>Item<\/th>\n<th>Value<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td>Repository<\/td>\n<td><a href=\"https:\/\/huggingface.co\/FastVideo\/FastH3-Omni-Comfy\">FastVideo\/FastH3-Omni-Comfy<\/a><\/td>\n<\/tr>\n<tr>\n<td>Family guide<\/td>\n<td><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/model-minimaxai-minimax-h3-en\/\">MiniMax-H3 guide (14 articles)<\/a><\/td>\n<\/tr>\n<tr>\n<td>Publisher guide<\/td>\n<td><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/publisher-minimax-en\/\">MiniMax: models and licenses<\/a><\/td>\n<\/tr>\n<tr>\n<td>Published<\/td>\n<td>2026-10-20<\/td>\n<\/tr>\n<tr>\n<td>License<\/td>\n<td>minimax-h3-community-license-agreement<\/td>\n<\/tr>\n<tr>\n<td>Formats<\/td>\n<td><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/format-safetensors-en\/\">safetensors<\/a><\/td>\n<\/tr>\n<tr>\n<td>Source type<\/td>\n<td>Primary source (the publisher itself)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<p><em>Values determined by this site&#8217;s code when the information was collected. Dates are JST.<\/em><\/p>\n<p><!-- \/lmw:facts --><\/p>\n<h2>Overview<\/h2>\n<p>FastVideo has released &#8220;FastVideo\/FastH3-Omni-Comfy&#8221;, a video and audio integrated generation model compatible with the ComfyUI environment. This model is a &#8220;Reference-to-Video&#8221; generation model that takes text prompts along with multiple ordered reference materials such as images, videos, and audio, and simultaneously generates video and time-synchronized audio matching those materials.<\/p>\n<p>The released models are built based on the base model &#8220;FastVideo\/FastVideo-FastH3-Omni-8-Step-V1&#8221; (root base model: MiniMaxAI\/MiniMax-H3), which supports high-speed inference in 8 steps. They are distributed in a single-file format (safetensors) that can be directly loaded and inferred in the ComfyUI environment. In addition to the BF16 version, which maintains standard precision, a memory-efficient INT8 ConvRot diffusion model is available as an option.<\/p>\n<h2>Specifications<\/h2>\n<p>The main specifications and configuration values confirmed from the release repository and materials for the base model &#8220;FastVideo\/FastVideo-FastH3-Omni-8-Step-V1&#8221; are as follows:<\/p>\n<ul>\n<li><strong>Base Model Information<\/strong>: The base model is FastVideo\/FastVideo-FastH3-Omni-8-Step-V1, and its underlying root base model is MiniMaxAI\/MiniMax-H3.<\/li>\n<li><strong>Distributed Models and File Formats<\/strong>: <\/li>\n<li><code>diffusion_models\/fasth3_omniref_8step_pruned_bf16.safetensors<\/code> (BF16 precision diffusion model) &#8211; <code>diffusion_models\/fasth3_omniref_8step_pruned_int8_convrot.safetensors<\/code> (INT8 ConvRot precision diffusion model)<\/li>\n<li><strong>Input Specifications<\/strong>:<\/li>\n<li>Accepts instructions via text prompts. &#8211; Reference inputs: Various materials can be specified, including images (up to 9), videos (up to 3), and audio (up to 3). Supports up to a total of 12 references in total. At least one image or video reference specification is required for generation. &#8211; Prompt tag specification: To instruct the order of reference materials, conditioning is performed by writing corresponding tags such as <code>&lt;image_1&gt;<\/code>, <code>&lt;video_1&gt;<\/code>, and <code>&lt;audio_1&gt;<\/code> within the text prompt. Note that if a video file is input as a reference, the audio track included in that video is also incorporated as reference data simultaneously.<\/li>\n<li><strong>Output Specifications<\/strong>:<\/li>\n<li>Video framerate: 24 FPS &#8211; Audio output: 32 kHz stereo audio. Outputted with time synchronized to the progression of the video. &#8211; Generation size and length limits: Depends on the execution environment used and available VRAM capacity (480p resolution is the default in the base model&#8217;s profile settings).<\/li>\n<li><strong>Recommended Inference Settings<\/strong>:<\/li>\n<li>Inference steps: 8 steps (completes generation with 8 transformer evaluations) &#8211; Guidance scale (guidance): 1.0 &#8211; Schedule shift values: Video shift is 12.0, Audio shift is 3.0<\/li>\n<li><strong>Required Component Configuration (for ComfyUI)<\/strong>:<\/li>\n<li>Text encoder: A Qwen3-VL 32B-based configuration including the Vision Tower required to interpret reference materials. Depending on the environment, select one of <code>qwen3vl_32b_minimax_h3_bf16.safetensors<\/code>, <code>qwen3vl_32b_minimax_h3_int8_convrot.safetensors<\/code>, or <code>qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors<\/code> (Blackwell generation hardware is required to use the acceleration features of the NVFP4 version). &#8211; Video VAE: Place <code>minimax_h3_video_vae_fp16.safetensors<\/code> or <code>minimax_h3_video_vae_int8_convrot.safetensors<\/code>. &#8211; Audio VAE: Place <code>minimax_h3_audio_vae_fp32.safetensors<\/code>, which is required to output synchronized audio. &#8211; Node requirements: A ComfyUI environment with native support for MiniMax-H3 and the use of the &#8220;MiniMax H3 Reference to Video&#8221; conditioning node are expected.<\/li>\n<li><strong>License Conditions<\/strong>: Complies with <code>minimax-h3-community-license-agreement<\/code> (license display: other). Please check the relevant license terms for usage conditions and restrictions.<\/li>\n<\/ul>\n<h2>Strengths and Use Cases<\/h2>\n<p>This model specializes in the &#8220;Reference-to-Video&#8221; task, where instead of relying solely on text prompt instructions, multiple media materials such as images, videos, and audio are loaded as reference data to generate video and synchronized audio reflecting those elements.<\/p>\n<p>Specifically, it is possible to ingest up to 9 images, up to 3 videos, and up to 3 audio files (up to 12 in total) as reference inputs. When a video reference is input, the audio track included in that video is also integrated and referenced. By using corresponding tags such as <code>&lt;image_1&gt;<\/code>, <code>&lt;video_1&gt;<\/code>, and <code>&lt;audio_1&gt;<\/code> within the text prompt to specify the order and relationship of the materials, it is possible to generate new videos and 32 kHz stereo audio while retaining the characteristics of existing materials.<\/p>\n<p>Therefore, it is suitable for the following use cases:<br \/>\n&#8211; Generating videos that develop new scenes and actions while consistently maintaining the visual expression of characters or specific objects, or acoustic elements<br \/>\n&#8211; Creating scene developments and cuts based on existing short video or audio materials<br \/>\n&#8211; Multimodal video and audio generation in an intuitive pipeline utilizing ComfyUI&#8217;s graphical node-based UI<\/p>\n<p>Furthermore, based on the performance of the base model FastVideo-FastH3-Omni-8-Step-V1, it can simultaneously generate video and synchronized audio with just 8 steps of transformer evaluation, making it suitable for workflows that want to balance generation speed and quality. In addition, the availability of the INT8 quantization format in addition to the BF16 format allows for flexible operation depending on the environment and computing resources used.<\/p>\n<h2>How It Differs from Similar Models<\/h2>\n<p>The <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/2026\/10\/06\/fasth3-v2-fp8-nvfp4-released\/\">FP8 and NVFP4 versions of the audio-inclusive video generation model &#8220;FastH3 V2&#8221;<\/a> and the <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/2026\/10\/06\/fasth3-trim-8-step\/\">8-step version of the audio-inclusive video generation model &#8220;FastH3 Trim&#8221; with reduced transformer blocks for lightweight operation<\/a>, which have been introduced on this site in the past, were primarily designed for Text-to-Video tasks generating videos with audio solely from text prompts, or focused on model structure miniaturization and acceleration.<\/p>\n<p>In contrast, this FastH3 Omni ComfyUI-supported model features &#8220;Reference-to-Video&#8221; functionality as a major characteristic, combining multiple reference materials such as images, videos, and audio alongside text for conditioning. It differs significantly from conventional models in that it not only generates video and audio from text, but also allows specifying concrete reference inputs (up to 12 items) to achieve consistent video expression and audio synchronization. Another major difference is that the ComfyUI-exclusive node configuration and separate placement of various VAEs and text encoders are organized, making it designed to be directly incorporated into a node-based visual operation environment.<\/p>\n<p><!-- lmw:hardware --><\/p>\n<h2>Hardware Requirements<\/h2>\n<p><strong>Estimated requirements (calculated by Local Model Watch)<\/strong><\/p>\n<div class=\"lmw-table-scroll\" tabindex=\"0\" style=\"overflow-x:auto;-webkit-overflow-scrolling:touch;max-width:100%;\">\n<table style=\"width:max-content;min-width:100%;border-collapse:collapse;\">\n<thead>\n<tr>\n<th>Your VRAM<\/th>\n<th>Quantization<\/th>\n<th>File size<\/th>\n<th>Est. memory needed<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td>80GB class (A100 \/ H100)<\/td>\n<td>Original precision(transformer_ref)<\/td>\n<td>65.3GB<\/td>\n<td>78.4GB<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<p><em>Memory estimates add a 20% runtime overhead (KV cache, etc.) to the actual size of the distributed files. Actual usage varies with context length, batch size and inference engine. These figures are computed by this site from file sizes, not published by the model&#8217;s authors. Compare with other models in our <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/vram-guide-en\/\">VRAM quick reference<\/a>. What the quantization names mean: <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/glossary-quantization-en\/\">glossary<\/a>.<\/em><\/p>\n<p><!-- \/lmw:hardware --><\/p>\n<p><!-- lmw:runnability --><\/p>\n<h2>Can You Run It Locally?<\/h2>\n<p>The publisher distributes this model as safetensors.<\/p>\n<p><strong>License \u2014 <code>minimax-h3-community-license-agreement<\/code> (Commercial use prohibited):<\/strong> MiniMax&#8217;s MiniMax H3 Community License. Use of the weights is limited to the EU, the UK, South Korea and the US (Sections II and V.4); elsewhere, including Japan, a separate license is required.<\/p>\n<p><em>Compiled by this site&#8217;s code from the published formats and the license field. License summaries are not legal advice \u2014 check the publisher&#8217;s original terms before relying on them.<\/em><\/p>\n<p><!-- \/lmw:runnability --><\/p>\n<p><!-- lmw:lab --><\/p>\n<h2>Our Own Measurements<\/h2>\n<h3>Measurements We Did Not Take<\/h3>\n<p>This model&#8217;s license (<code>minimax-h3-community-license-agreement<\/code>) does not permit commercial use. Because this site carries advertising, we did not run the model (no CPU run, answers, quantization comparison, conversion or generation).<\/p>\n<p><!-- \/lmw:lab --><\/p>\n<p><!-- lmw:files --><\/p>\n<h2>Distributed Files<\/h2>\n<p><em>Weight files published in <a href=\"https:\/\/huggingface.co\/FastVideo\/FastH3-Omni-Comfy\/tree\/main\">FastVideo\/FastH3-Omni-Comfy<\/a>, listed by this site from the Hugging Face API. Sizes are the actual file sizes.<\/em><\/p>\n<div class=\"lmw-table-scroll\" tabindex=\"0\" style=\"overflow-x:auto;-webkit-overflow-scrolling:touch;max-width:100%;\">\n<table style=\"width:max-content;min-width:100%;border-collapse:collapse;\">\n<thead>\n<tr>\n<th>File<\/th>\n<th>Size<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td><code>diffusion_models\/fasth3_omniref_8step_pruned_bf16.safetensors<\/code><\/td>\n<td>44.24GB<\/td>\n<\/tr>\n<tr>\n<td><code>diffusion_models\/fasth3_omniref_8step_pruned_int8_convrot.safetensors<\/code><\/td>\n<td>22.21GB<\/td>\n<\/tr>\n<tr>\n<td><code>text_encoders\/qwen3vl_32b_minimax_h3_bf16.safetensors<\/code><\/td>\n<td>51.51GB<\/td>\n<\/tr>\n<tr>\n<td><code>text_encoders\/qwen3vl_32b_minimax_h3_int8_convrot.safetensors<\/code><\/td>\n<td>27.14GB<\/td>\n<\/tr>\n<tr>\n<td><code>text_encoders\/qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors<\/code><\/td>\n<td>15.69GB<\/td>\n<\/tr>\n<tr>\n<td><code>vae\/minimax_h3_audio_vae_fp32.safetensors<\/code><\/td>\n<td>605MB<\/td>\n<\/tr>\n<tr>\n<td><code>vae\/minimax_h3_video_vae_fp16.safetensors<\/code><\/td>\n<td>5.21GB<\/td>\n<\/tr>\n<tr>\n<td><code>vae\/minimax_h3_video_vae_int8_convrot.safetensors<\/code><\/td>\n<td>2.81GB<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n<p><!-- \/lmw:files --><\/p>\n<h2>How to Get It<\/h2>\n<p>These models are distributed in safetensors format and can be obtained from the Hugging Face repository <code>FastVideo\/FastH3-Omni-Comfy<\/code>. Prior terms of use agreement procedures (Gated models) are not required to obtain them, but the <code>minimax-h3-community-license-agreement<\/code> license applies to their distribution.<\/p>\n<p>When introducing them, use the latest version of ComfyUI including native support for MiniMax-H3, and place the files from the repository into their corresponding folders. The specific directory structure for placement is as follows:<\/p>\n<ol>\n<li>\n<p><strong>Diffusion Models<\/strong><br \/>\n   Select one of the following safetensors files and place it in the <code>ComfyUI\/models\/diffusion_models\/<\/code> directory. &#8211; <code>fasth3_omniref_8step_pruned_bf16.safetensors<\/code> (BF16 precision model) &#8211; <code>fasth3_omniref_8step_pruned_int8_convrot.safetensors<\/code> (INT8 ConvRot precision model)<\/p>\n<\/li>\n<li>\n<p><strong>Text Encoders<\/strong><br \/>\n   Select from the <code>text_encoders\/<\/code> folder in the repository and place them in the <code>ComfyUI\/models\/text_encoders\/<\/code> directory. Includes the Vision Tower required to process reference materials. &#8211; <code>qwen3vl_32b_minimax_h3_bf16.safetensors<\/code> &#8211; <code>qwen3vl_32b_minimax_h3_int8_convrot.safetensors<\/code> &#8211; <code>qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors<\/code> (for NVFP4 acceleration on Blackwell hardware)<\/p>\n<\/li>\n<li>\n<p><strong>VAE Models (VAE)<\/strong><br \/>\n   Obtain from the <code>vae\/<\/code> folder in the repository and place them in the <code>ComfyUI\/models\/vae\/<\/code> directory. Placing the audio VAE is mandatory to generate synchronized audio. &#8211; Video VAE: <code>minimax_h3_video_vae_fp16.safetensors<\/code> or <code>minimax_h3_video_vae_int8_convrot.safetensors<\/code> &#8211; Audio VAE: <code>minimax_h3_audio_vae_fp32.safetensors<\/code><\/p>\n<\/li>\n<\/ol>\n<p>During execution, use the &#8220;MiniMax H3 Reference to Video&#8221; conditioning node on ComfyUI, appropriately connect each component of the text encoder, video VAE, and audio VAE, and perform inference.<\/p>\n<p><!-- lmw:related --><\/p>\n<h2>Related Articles<\/h2>\n<ul>\n<li><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/11\/17\/fastvideo-fasth3-trim-comfy-released\/\">FastVideo-FastH3-Trim-Comfy Text-to-Video and Audio Model: File List<\/a><\/li>\n<li><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/11\/17\/fasth3-omni-fp8-comfy-released-2\/\">FastH3-Omni-FP8-Comfy Audio-Synced Video Generation Model: 32GB+ VRAM<\/a><\/li>\n<li><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/11\/17\/fasth3-omni-nvfp4-qad-comfy-released\/\">FastH3-Omni-NVFP4-QAD-Comfy: 16GB+ VRAM, File List<\/a><\/li>\n<li><a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/2026\/11\/17\/fasth3-omniref-comfy-released\/\">FastH3-OmniRef-Comfy: File List<\/a><\/li>\n<\/ul>\n<p><!-- \/lmw:related --><\/p>\n<p><!-- lmw:next-steps --><\/p>\n<h2>What to Read Next<\/h2>\n<ul>\n<li><strong>Find models by VRAM<\/strong> (This model needs at least 80GB) \u2192 <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/vram-guide-en\/\">VRAM quick reference<\/a><\/li>\n<li><strong>Explore the same model family<\/strong> \u2192 <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/model-minimaxai-minimax-h3-en\/\">MiniMax-H3 family overview (14 articles, 4 converted builds)<\/a><\/li>\n<li><strong>Formats this model is available in<\/strong> \u2192 <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/format-safetensors-en\/\">Safetensors format guide and models<\/a><\/li>\n<li><strong>Learn about the publisher<\/strong> \u2192 <a href=\"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/publisher-minimax-en\/\">MiniMax: models, licenses and articles<\/a><\/li>\n<\/ul>\n<p><!-- \/lmw:next-steps --><\/p>\n<h2>Sources<\/h2>\n<ul>\n<li><a href=\"https:\/\/huggingface.co\/FastVideo\/FastH3-Omni-Comfy\">https:\/\/huggingface.co\/FastVideo\/FastH3-Omni-Comfy<\/a><\/li>\n<li><a href=\"https:\/\/huggingface.co\/FastVideo\/FastVideo-FastH3-Omni-8-Step-V1\">https:\/\/huggingface.co\/FastVideo\/FastVideo-FastH3-Omni-8-Step-V1<\/a><\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"<p>FastVideo has released FastH3-Omni-Comfy, a ComfyUI-compatible video and audio generation model supporting Reference-to-Video tasks.<\/p>\n","protected":false},"author":1,"featured_media":12743,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[203],"tags":[677,3877,3031,730,209,1547],"class_list":["post-12744","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-image-video-and-audio","tag-comfyui-en","tag-fasth3-omni-comfy-en","tag-fastvideo-en","tag-minimax-h3-en","tag-safetensors-en","tag-verified"],"lang":"en","translations":{"en":12744,"ja":12742},"_links":{"self":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/12744","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/comments?post=12744"}],"version-history":[{"count":4,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/12744\/revisions"}],"predecessor-version":[{"id":18253,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/posts\/12744\/revisions\/18253"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media\/12743"}],"wp:attachment":[{"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/media?parent=12744"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/categories?post=12744"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/localmodelwatch.tsuchitsuchi.com\/en\/wp-json\/wp\/v2\/tags?post=12744"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}