[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-94902":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":10,"language":11,"languages":10,"totalLinesOfCode":10,"stars":12,"forks":13,"watchers":14,"openIssues":15,"contributorsCount":16,"subscribersCount":16,"size":16,"stars1d":16,"stars7d":17,"stars30d":17,"stars90d":16,"forks30d":16,"starsTrendScore":18,"compositeScore":19,"rankGlobal":10,"rankLanguage":10,"license":20,"archived":21,"fork":21,"defaultBranch":22,"hasWiki":21,"hasPages":21,"topics":23,"createdAt":10,"pushedAt":10,"updatedAt":24,"readmeContent":25,"aiSummary":26,"trendingCount":16,"starSnapshotCount":16,"syncStatus":14,"lastSyncTime":27,"discoverSource":28},94902,"ComfyUI-ALLinONE-MinimaxH3","LeonQ8\u002FComfyUI-ALLinONE-MinimaxH3","LeonQ8","All-in-one MiniMax H3 node for ComfyUI - T2V, I2V, R2V, audio drive, keyframes, extend & Motion Context chains in a single node.","",null,"JavaScript",247,30,2,1,0,146,6,74.47,"GNU General Public License v3.0",false,"main",[],"2026-08-24 04:01:22","# ComfyUI ALL-in-ONE MiniMax H3\n\n![Status: Beta](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fstatus-beta-orange)\n![License: GPL-3.0](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Flicense-GPL--3.0-blue)\n\nOne node. The whole MiniMax H3 video pipeline.\n\nNo node graph to build, no wires to connect, no hunting through twelve custom node packs to figure out which workflow is the right one. Pick a mode, drop in your prompt or references, hit **Generate** — the node does the rest.\n\n![ALL in ONE MiniMaxH3 — T2V tab](assets\u002Ft2v_main.png)\n\n## Modes\n\n| Mode | What it does |\n|------|--------------|\n| **Image** | Still images with H3: text to image, image edit, or reference mix (up to 9 references) via H3 Studio |\n| **T2V** | Text to video with native audio (fl2va model) |\n| **I2V** | Animate a start frame, optionally morph to an end frame |\n| **R2V** | Reference images \u002F videos \u002F audio drive the clip (ref2va model) |\n| **Audio Drive** | Your audio track is the soundtrack, and it drives mouth movement (lip sync) |\n| **Keyframes** | Pin still images at arbitrary frame positions |\n| **Extend** | Continue an existing video seamlessly |\n| **Chain** | Multi-clip continuation with H3 Motion Context (latent path, no re-encode) |\n| **Upscale** | RTX\u002FSeed2VR Video Super Resolution hook |\n\n## Screenshots\n\n**History** — searchable, with prompt reuse and per-entry preview.\n\n![History](assets\u002Fhistory.png)\n\n**Library** — every output in one place: inline preview, favorites, open-folder, delete, RTX upscale hook.\n\n![Library](assets\u002Flibrary.png)\n\n**Settings** — theme accent, sounds, models.\n\n![Settings](assets\u002Fsettings.png)\n\n## Requirements\n\n### Models\n\nOfficial MiniMax H3 files from [Comfy-Org\u002FMiniMax-H3](https:\u002F\u002Fhuggingface.co\u002FComfy-Org\u002FMiniMax-H3), placed in your standard `ComfyUI\u002Fmodels\u002F` folders:\n\n| File | Folder |\n|------|--------|\n| `minimax_h3_fl2va_pruned_int8_convrot.safetensors` | `diffusion_models\u002F` |\n| `minimax_h3_ref2va_pruned_int8_convrot.safetensors` | `diffusion_models\u002F` |\n| `qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors` | `text_encoders\u002F` |\n| `minimax_h3_video_vae_fp16.safetensors` | `vae\u002F` |\n| `minimax_h3_audio_vae_fp32.safetensors` | `vae\u002F` |\n\n### Custom nodes\n\n**T2V, I2V and R2V need nothing extra** — every node they use (H3 conditioning, sigma shift, samplers, video\u002Faudio decode, video save) ships with a recent ComfyUI. The other modes and presets use a few community packs — install only the ones you use, via ComfyUI-Manager (search by pack name), then fully restart ComfyUI and hard-refresh the browser (`Ctrl+F5`).\n\n**Per mode**\n\n| Mode | Packs you need |\n|------|----------------|\n| T2V \u002F I2V \u002F R2V | — (ComfyUI core only) |\n| Audio Drive | [comfyui-vrgamedevgirl](https:\u002F\u002Fgithub.com\u002Fvrgamegirl19\u002Fcomfyui-vrgamedevgirl) |\n| Keyframes | [ComfyUI-H3-Motion-Context-MultiRef](https:\u002F\u002Fgithub.com\u002Fseitanism\u002FComfyUI-H3-Motion-Context-MultiRef) |\n| Extend | [ComfyUI-H3-Motion-Context-MultiRef](https:\u002F\u002Fgithub.com\u002Fseitanism\u002FComfyUI-H3-Motion-Context-MultiRef) |\n| Chain | [ComfyUI-H3-Motion-Context-MultiRef](https:\u002F\u002Fgithub.com\u002Fseitanism\u002FComfyUI-H3-Motion-Context-MultiRef) |\n| Upscale — RTX VSR | [Nvidia_RTX_Nodes_ComfyUI](https:\u002F\u002Fgithub.com\u002FComfy-Org\u002FNvidia_RTX_Nodes_ComfyUI) (NVIDIA RTX GPUs only) |\n| Upscale — SeedVR2 | [ComfyUI-SeedVR2_VideoUpscaler](https:\u002F\u002Fgithub.com\u002Fnumz\u002FComfyUI-SeedVR2_VideoUpscaler) |\n| Image | [ComfyUI-MiniMax-H3-Studio](https:\u002F\u002Fgithub.com\u002Fthaakeno\u002FComfyUI-MiniMax-H3-Studio) |\n\n**Per quality preset** (Settings → Quality)\n\n| Preset | Packs you need |\n|--------|----------------|\n| Turbo | [ComfyUI-MiniMax-H3-Turbo](https:\u002F\u002Fgithub.com\u002FLarryvrh\u002FComfyUI-MiniMax-H3-Turbo) + a Turbo LoRA (below) |\n| Speed | [ComfyUI-SolAttn_triton](https:\u002F\u002Fgithub.com\u002Fkijai\u002FComfyUI-SolAttn_triton) + [ComfyUI-MiniMaxH3-Cache](https:\u002F\u002Fgithub.com\u002Flihaoyun6\u002FComfyUI-MiniMaxH3-Cache) |\n| Balanced | [ComfyUI-SolAttn_triton](https:\u002F\u002Fgithub.com\u002Fkijai\u002FComfyUI-SolAttn_triton) |\n| High | [ComfyUI-KJNodes](https:\u002F\u002Fgithub.com\u002Fkijai\u002FComfyUI-KJNodes) (SageAttention) |\n| Native | — (ComfyUI core only) |\n\nEach accelerator also has an on\u002Foff chip under the Quality dropdown (SolAttn \u002F H3 Cache \u002F SageAttn) — flip them for any mix; the preset label switches to **Custom**. Accelerators that are switched off are not even written into the workflow, so their packs don't need to be installed.\n\n**Preview without saving** (auto-save off): [ComfyUI-VideoHelperSuite](https:\u002F\u002Fgithub.com\u002FKosinkadink\u002FComfyUI-VideoHelperSuite)\n\n**Live Preview** (the toggle under the video): [ComfyUI-MiniMax-H3-Studio](https:\u002F\u002Fgithub.com\u002Fthaakeno\u002FComfyUI-MiniMax-H3-Studio) plus `taeh3.safetensors` in `ComfyUI\u002Fmodels\u002Fvae_approx\u002F` ([Kijai\u002FMiniMax-H3-TAE](https:\u002F\u002Fhuggingface.co\u002FKijai\u002FMiniMax-H3-TAE)). Shows the video while it samples; slows generation a little. Works in every video mode, not with the Turbo preset or Image mode. Your copy can sit in a subfolder of `vae_approx`, pick it under Settings: Live Preview decoder. Tested on ComfyUI 0.32.\n\n**Image mode prompts**: they follow the H3 Studio shape, a `summary:` line with the goal and a `detailed_description:` with the full scene. Name your references `@Image1`, `@Image2` and give each one a clear job (identity, pose, style, outfit). Edits are a semantic regeneration of the source image, not pixel inpainting, so describe what changes instead of expecting a perfect cutout. The Discover tab ships with Text to image, Image edit and Reference mix templates.\n\n**Image mode models**: besides your usual H3 files, H3 Studio's prompt machinery wants two small Qwen3.5 models in `ComfyUI\u002Fmodels\u002Ftext_encoders\u002F`:\n\n| Model | Download |\n|-------|----------|\n| `qwen3.5_2b_bf16.safetensors` | [Comfy-Org\u002FQwen3.5](https:\u002F\u002Fhuggingface.co\u002FComfy-Org\u002FQwen3.5\u002Fresolve\u002Fmain\u002Ftext_encoders\u002Fqwen3.5_2b_bf16.safetensors) |\n| `qwen3.5_4b_bf16.safetensors` | [Comfy-Org\u002FQwen3.5](https:\u002F\u002Fhuggingface.co\u002FComfy-Org\u002FQwen3.5\u002Fresolve\u002Fmain\u002Ftext_encoders\u002Fqwen3.5_4b_bf16.safetensors) |\n\n**LightX LoRAs for Image mode** (only the LightX sampling profiles need them, Base profiles need nothing). Drop the file into `ComfyUI\u002Fmodels\u002Floras\u002F`, the node checks for it before generating:\n\n| Profile | LoRA file |\n|---------|-----------|\n| LightX v1.0 FL2VA 8 steps | [official full](https:\u002F\u002Fhuggingface.co\u002Flightx2v\u002FMinimax-h3-Turbo\u002Fresolve\u002Fmain\u002Fminimax_h3_fl2v_turbo_8step_v1.0_comfyui_bf16.safetensors) or [Kijai rank 24](https:\u002F\u002Fhuggingface.co\u002FKijai\u002FMiniMax-H3_comfy\u002Fresolve\u002Fmain\u002Floras\u002Fminimax_h3_fl2v_lightx2v_turbo_8step_v1.0_resized_avg_rank_24_bf16.safetensors) |\n| LightX v1.0 FL2VA 4 steps | [Kijai rank 31](https:\u002F\u002Fhuggingface.co\u002FKijai\u002FMiniMax-H3_comfy\u002Fresolve\u002Fmain\u002Floras\u002Fminimax_h3_fl2v_lightx2v_turbo_4step_v1.0_768p_resized_avg_rank_31_bf16.safetensors) |\n| LightX v0.1 ER-SDE \u002F SA-Solver | [Kijai rank 21](https:\u002F\u002Fhuggingface.co\u002FKijai\u002FMiniMax-H3_comfy\u002Fresolve\u002Fmain\u002Floras\u002Fminimax_h3_fl2v_lightx2v_turbo_4step_v0.1_comfy_resized_avg_rank_21_bf16.safetensors) |\n| LightX v0.1 REF2V (Reference Mix) | [Kijai rank 20](https:\u002F\u002Fhuggingface.co\u002FKijai\u002FMiniMax-H3_comfy\u002Fresolve\u002Fmain\u002Floras\u002Fminimax_h3_ref2v_lightx2v_turbo_4step_v0.1_resized_avg_rank_20_bf16.safetensors) |\n\n**Turbo LoRA** (for the Turbo preset): download `minimax_h3_turbo_v4_step600_ema.safetensors` from [larryvrh\u002FMiniMax-H3-Turbo-Lora](https:\u002F\u002Fhuggingface.co\u002Flarryvrh\u002FMiniMax-H3-Turbo-Lora) into `ComfyUI\u002Fmodels\u002Floras\u002F`.\n\n> **Seeing \"Node not found\"?** That's a missing pack from the tables above. The two most common:\n> - `Audio Drive` node → install **comfyui-vrgamedevgirl**\n> - Extend \u002F Chain \u002F Keyframes nodes → install **ComfyUI-H3-Motion-Context-MultiRef**\n>\n> Install via ComfyUI-Manager, restart ComfyUI completely, then hard-refresh the browser.\n\nExact tested versions of everything are in **[COMPATIBILITY.md](COMPATIBILITY.md)** — check that file first if something breaks after you update ComfyUI or a pack.\n\n## Installation\n\n```bash\n# inside ComfyUI\u002Fcustom_nodes\u002F\ngit clone https:\u002F\u002Fgithub.com\u002FLeonQ8\u002FComfyUI-ALLinONE-MinimaxH3.git\n```\n\nRestart ComfyUI, then double-click the canvas and search for **ALL in ONE MiniMaxH3**.\n\n## Compatibility\n\nI develop and test against a pinned stack (ComfyUI version, custom node commits, model files). It's all listed in **[COMPATIBILITY.md](COMPATIBILITY.md)**, if a render fails after you updated something, start there.\n\n## Credits\n\n- The \"one node\" idea and UI approach: Ján — [one-node-flux-2-klein](https:\u002F\u002Fgithub.com\u002Fyanokusnir-ai\u002Fone-node-flux-2-klein) and [one-node-gemma-4](https:\u002F\u002Fgithub.com\u002Fyanokusnir-ai\u002Fone-node-gemma-4)\n- Chain \u002F Keyframes \u002F Extend wiring: [ComfyUI-H3-Motion-Context-MultiRef](https:\u002F\u002Fgithub.com\u002Fseitanism\u002FComfyUI-H3-Motion-Context-MultiRef) by seitanism\n- Base graphs: the official MiniMax H3 native workflows from Comfy-Org\n- Turbo preset: ComfyUI-MiniMax-H3-Turbo pack\n- Image mode: [ComfyUI-MiniMax-H3-Studio](https:\u002F\u002Fgithub.com\u002Fthaakeno\u002FComfyUI-MiniMax-H3-Studio) by thaakeno\n\n## Support\n\nThis node is in **beta** — if something breaks, please [open an issue](https:\u002F\u002Fgithub.com\u002FLeonQ8\u002FComfyUI-ALLinONE-MinimaxH3\u002Fissues), it's the fastest way to get it fixed.\n\nIf you like this node and it saves you a few hours of graph surgery, a coffee is always appreciated.\u003C3\n\n\u003Ca href=\"https:\u002F\u002Fko-fi.com\u002Fleonq8\" target=\"_blank\">\u003Cimg height=\"36\" style=\"border:0px;height:36px;\" src=\"https:\u002F\u002Fstorage.ko-fi.com\u002Fcdn\u002Fkofi5.png?v=3\" border=\"0\" alt=\"Buy Me a Coffee at ko-fi.com\" \u002F>\u003C\u002Fa>\n\n## License\n\nGPL-3.0 — see [LICENSE](LICENSE).\n","这是一个为 ComfyUI 设计的集成化 MiniMax H3 视频生成节点，将文本\u002F图像\u002F音频驱动视频（T2V\u002FI2V\u002FR2V）、唇形同步、关键帧控制、视频续写与扩展、Motion Context 链式生成及超分等功能封装于单个节点中。核心特点是免布线、模式化操作、内置完整 H3 工作流（含 FL2VA\u002FREF2VA 模型支持、多参考混合、音频驱动 VAE 解码等），并提供历史记录、媒体库与统一设置界面。适用于需要高效迭代视频生成效果的研究者、AIGC 内容创作者及 ComfyUI 中高级用户，尤其适合快速验证创意、批量生成短视频或构建可复用的视频生成流程。","2026-08-17 02:30:13","CREATED_QUERY"]