--- title: Minimax-H3-ComfyUI canonical_url: "https://www.modelscope.cn/models/Alissonerdx/Minimax-H3-ComfyUI" md_url: "https://www.modelscope.cn/models/Alissonerdx/Minimax-H3-ComfyUI.md" repository: Alissonerdx/Minimax-H3-ComfyUI last_updated: 2026-11-11 license: other base_model: - MiniMaxAI/MiniMax-H3 base_model_relation: adapter parameters: 12.3B tensor_type: - F32 - BF16 - F16 library_name: - lora - safetensors - pytorch frameworks: - pytorch downloads: 3014 stars: 7 tags: - lora - video - comfyui - minimax-h3 --- # Minimax-H3-ComfyUI > Minimax-H3-ComfyUI - Alissonerdx 在 ModelScope 开源的模型。MiniMax H3 LoRAs for ComfyUI Alissonerdx/Minimax-H3-ComfyUI 是 ModelScope 魔搭社区上的 12.3B 参数机器学习模型，采用 other 许可，基于 MiniMaxAI/MiniMax-H3 构建。 - **Repository**: Alissonerdx/Minimax-H3-ComfyUI - **License**: other - **Parameters**: 12.3B - **Base model**: MiniMaxAI/MiniMax-H3 - **Tags**: lora, video, comfyui, minimax-h3 - **Downloads**: 3014 - **Stars**: 7 - **Last updated**: 2026-11-11 Source: https://www.modelscope.cn/models/Alissonerdx/Minimax-H3-ComfyUI --- # MiniMax H3 LoRAs for ComfyUI Experimental LoRAs for [MiniMax H3](https://huggingface.co/MiniMaxAI/MiniMax-H3), packaged for ComfyUI and tested mainly with the [`ref2va` weights](https://huggingface.co/Comfy-Org/MiniMax-H3/tree/main/diffusion_models). These are research models, not finished tools. Results vary substantially with the source clip, resolution, prompt and reference images. ## Models | Model | Purpose | Trigger | Documentation | Workflow | |---|---|---|---|---| | [`minimax_h3_lms_v1.0_r64`](loras/minimax_h3_lms_v1.0_r64.safetensors) | Sharpness/detail enhancement using an aligned source-video guide | Fixed enhancement caption | [LMS / Sharpness guide](docs/lms.md) | [`minimax_h3_lms_workflow.json`](workflows/minimax_h3_lms_workflow.json) | | [`minimax_h3_style_transfer_v1.0_r64`](loras/minimax_h3_style_transfer_v1.0_r64.safetensors) | Transfer a visual style from an image or text description onto a video | `style_transfer:` | [Style Transfer guide](docs/style-transfer.md) | [`minimax_h3_style_transfer_workflow.json`](workflows/minimax_h3_style_transfer_workflow.json) | | [`minimax_h3_vfx_edit_v1.0_r128`](loras/minimax_h3_vfx_edit_v1.0_r128.safetensors) | General prompt-directed VFX editing with an aligned source video | `vfx_edit:` | [VFX Edit guide](docs/vfx-edit.md) | [`minimax_h3_vfxedit_wokflow.json`](workflows/minimax_h3_vfxedit_wokflow.json) | | [`minimax_h3_vfx_edit_v1.0_r128_ffp`](loras/minimax_h3_vfx_edit_v1.0_r128_ffp.safetensors) | Higher-resolution VFX Edit version with experimental first-frame propagation | `vfx_edit:` | [VFX Edit and FFP guide](docs/vfx-edit.md) | [`minimax_h3_vfxedit_wokflow.json`](workflows/minimax_h3_vfxedit_wokflow.json) | | [`minimax_h3_head_swap_v1.0_r32`](loras/minimax_h3_head_swap_v1.0_r32.safetensors) | Experimental reference-image head replacement over an aligned source video | `head_swap:` | [Head Swap guide and prompt template](docs/head-swap.md) | [`minimax_h3_head_swap_workflow.json`](workflows/minimax_h3_head_swap_workflow.json) | | [Character Swap Preprocessor Map Fro 0.99](loras/minimax_h3_character_swap_v1.0_preprocessor_map_fro099.safetensors) | Replace color-marked subjects using normal maps or DWPose guides on Ref2VA. Partial H3 Image support | `character_swap:` | [Setup and compatibility](docs/character-swap-preprocessor-map.md) | [Video](workflows/minimax_h3_character_swap_preprocessor_map_workflow_v1.0.json) / [Image (experimental)](workflows/minimax_h3_image_character_swap_preprocessor_map_workflow_v1.0.json) | | [`minimax_h3_character_swap_v1.1_ref2va_fro`](loras/minimax_h3_character_swap_v1.1_ref2va_fro.safetensors) **(recommended)** | Replace the main guide subject with the character in a reference image while following the guide performance; 80 MB SVD extraction | `body_swap:` | [Character Swap guide, prompt template and examples](docs/character-swap.md) | [Shot Planner](workflows/minimax_h3_character_swap_workflow_v1.1.json) / [native nodes](workflows/minimax_h3_character_swap_workflow_native_nodes_v1.0.json) | | [`minimax_h3_character_swap_v1.0_ref2va_r64_gl`](loras/minimax_h3_character_swap_v1.0_ref2va_r64_gl.safetensors) | Previous Character Swap release (rank 64) | `body_swap:` | [Character Swap guide](docs/character-swap.md) | [Shot Planner](workflows/minimax_h3_character_swap_workflow_v1.0.json) / [native nodes](workflows/minimax_h3_character_swap_workflow_native_nodes_v1.0.json) | | [`minimax_h3_motion_transfer_v1.0_ref2va`](loras/minimax_h3_motion_transfer_v1.0_ref2va.safetensors) **(new)** | Motion Transfer: the character in the reference performs the guide video's motion; carries [ostris' training adapter V2](https://huggingface.co/ostris/minimax_h3_training_adapter) merged in (185 MB) | `motion_transfer:` | [Native Guide Latents](workflows/minimax_h3_motion_transfer_workflow_native_nodes_v1.0.json) | [Planner Guide Latents](workflows/minimax_h3_motion_transfer_workflow_planner_v1.0.json) | | [`minimax_h3_motion_transfer_tara_v1.0_ref2va`](loras/minimax_h3_motion_transfer_tara_v1.0_ref2va.safetensors) **(new)** | Motion Transfer with the guide as a native video reference, aligned frame by frame by TARA (`BFSMiniMaxH3NativeTemporal`); training adapter V2 merged in (126 MB) | `motion_transfer:` | [Native TARA](workflows/minimax_h3_motion_transfer_tara_workflow_native_nodes_v1.0.json) | [Planner TARA](workflows/minimax_h3_motion_transfer_tara_workflow_planner_v1.0.json) | | [`minimax_h3_motion_transfer_ffp_v1.0_ref2va`](loras/minimax_h3_motion_transfer_ffp_v1.0_ref2va.safetensors) **(new)** | Motion Transfer FFP: edit the first frame (for example replace the character) and the edit follows the whole video. VFX Edit FFP (rank 128) × 1.0 + SolarRenderer control × 0.5, SVD-compressed to 99% of the Frobenius energy (230 MB) | `vfx_edit:` | [Henge](https://github.com/alisson-anjos/Henge) (Motion Transfer · First Frame) | [Licenses](LICENSES.md#merges-with-solarrenderer) | | [`solarrenderer_control_comfy_fro_dyn99`](loras/solarrenderer/solarrenderer_control_comfy_fro_dyn99.safetensors) / [full rank](loras/solarrenderer/solarrenderer_control_comfy.safetensors) | [SolarRenderer](https://huggingface.co/Enderfga/SolarRenderer)'s control LoRA converted to ComfyUI H3 keys; the `_fro_dyn99` copy keeps 99% of each module's Frobenius energy (757 MB instead of 5.9 GB). Used alone by Henge's Motion Transfer SolarRenderer FFP, on SolarRenderer's own conditioning | SolarRenderer six-part prompt | [Henge](https://github.com/alisson-anjos/Henge) (Motion Transfer SolarRenderer FFP) | [Licenses](LICENSES.md#merges-with-solarrenderer) | | [`solarrenderer_4step_comfy_fro_dyn99`](loras/solarrenderer/solarrenderer_4step_comfy_fro_dyn99.safetensors) / [full rank](loras/solarrenderer/solarrenderer_4step_comfy.safetensors) | SolarRenderer's four-step LoRA converted to ComfyUI H3 keys; `_fro_dyn99` at 99% of the Frobenius energy (1.4 GB instead of 5.9 GB). Load it before the control LoRA | — | [Henge](https://github.com/alisson-anjos/Henge) | [Licenses](LICENSES.md#merges-with-solarrenderer) | | [`minimax_h3_live_wallpaper_v1.0_ref2va_r32`](loras/minimax_h3_live_wallpaper_v1.0_ref2va_r32.safetensors) | Animate one to three reference images; often more faithful to facial features, but with less predictable camera behavior | `live_wallpaper:` | [Live Wallpaper guide](docs/live-wallpaper.md) | [`minimax_h3_live_wallpaper_workflow.json`](workflows/minimax_h3_live_wallpaper_workflow.json) | | [`minimax_h3_live_wallpaper_v1.0_ref2va_r64`](loras/minimax_h3_live_wallpaper_v1.0_ref2va_r64.safetensors) | Camera-aware continuation of R32 for prompt-directed camera behavior and temporal reference frames | `live_wallpaper:` | [Live Wallpaper guide](docs/live-wallpaper.md) | [`minimax_h3_live_wallpaper_workflow.json`](workflows/minimax_h3_live_wallpaper_workflow.json) | ## Motion Transfer and TARA **Four downloadable workflows:** [Guide Latents / native](workflows/minimax_h3_motion_transfer_workflow_native_nodes_v1.0.json), [Guide Latents / Planner](workflows/minimax_h3_motion_transfer_workflow_planner_v1.0.json), [TARA / native](workflows/minimax_h3_motion_transfer_tara_workflow_native_nodes_v1.0.json), [TARA / Planner](workflows/minimax_h3_motion_transfer_tara_workflow_planner_v1.0.json). **References are 1..N pictures with roles assigned by the prompt, not a required person/background pair.** Several images can describe one identity, multiple characters, outfits, objects, scenery or style. A separate background image is optional. See the [Motion Transfer guide](docs/motion-transfer.md) for model setup, multiple-reference examples and the **modified temporal Comfy Kitchen kernel** required for TARA's fused one-pass Kitchen path. The TARA workflows default to Flash partitions, which does not require that extension. **Training adapter V2.** The Motion Transfer LoRAs were trained on top of [ostris' training adapter V2](https://huggingface.co/ostris/minimax_h3_training_adapter) (`minimax_h3_ref2va_training_adapter_v2.safetensors`) kept frozen. Since 2026-11-02 the two files above carry it: LoRA x1 + adapter x1 merged by exact rank concat, then per-module SVD keeping 99% of the Frobenius energy (185 MB and 126 MB instead of 620 MB + 310 MB). Load only turbo + motion LoRA; **do not load the adapter again**. In same-seed tests they matched the full LoRA + separate adapter, including TARA's frame sync across camera cuts. The previous files (without the adapter) are kept in [`loras/base_loras_ignore/`](loras/base_loras_ignore). **TARA-H3 — Temporally Aligned Reference Attention.** Native reference conditioning alone did not reliably reproduce the timing of a source performance in H3, while aligned guide latents (Add Guide) copy it frame by frame. TARA keeps the motion video as native reference tokens, translates its temporal RoPE coordinates onto the generated video's clock, and adds a bounded temporal penalty only to generated-video queries attending to motion-reference keys; image references keep their native coordinates and the attention change adds no learned layers. Motion Transfer TARA v1.0 was trained for this method. Node: `BFSMiniMaxH3NativeTemporal` in [ComfyUI-BFSNodes](https://github.com/alisson-anjos/ComfyUI-BFSNodes). Technical report (v0.1.2, experimental, not peer reviewed): [TARA-H3: Temporally Aligned Reference Attention for Native Video Conditioning](https://github.com/alisson-anjos/ai-toolkit/tree/tara-h3-v0.1.2/research/tara-h3). All of these run in [Henge](https://github.com/alisson-anjos/Henge), the open-source studio built on them. ## Latent Upscaler [`h3_upscaler_lms_v0.1`](latent_upscaler/h3_upscaler_lms_v0.1.safetensors) is the recommended FP16 release of an experimental latent-upscaler checkpoint. It was trained for 2,000 steps with the same sharpness dataset used for the LMS LoRA and initialized from the [LBH-123-AI/Minimax_h3_latent_Upscaler](https://huggingface.co/LBH-123-AI/Minimax_h3_latent_Upscaler) project and its [`minimax_h3_latent_upscaler_3d_fp16.safetensors`](https://huggingface.co/LBH-123-AI/Minimax_h3_latent_Upscaler/blob/main/minimax_h3_latent_upscaler_3d_fp16.safetensors) checkpoint. See the [Latent Upscaler guide](docs/latent-upscaler.md) for setup, scope and limitations. ## Still images with Pruna P-H3-Image (guide-latent LoRAs) [`PrunaImageH3WorkaroundGuideLatents.json`](workflows/PrunaImageH3WorkaroundGuideLatents.json) is a workaround to use [PrunaAI's P-H3-Image](https://huggingface.co/PrunaAI/P-H3-Image-Comfy) with `MiniMaxH3AddGuide`, so LoRAs trained on guide latents can edit a single image. `PH3ImageConditioning` with no images gives the one-frame H3 latent, `MiniMaxH3AddGuide` adds the image to edit as the guide at `frame_idx = 0`, `MiniMaxH3ReferenceToVideo` (length 5) carries the prompt and the reference picture, and the P-H3-Image VAE decodes the single frame. The model stack is the H3 `ref2va` weights, the Pika turbo LoRA, the P-H3-Image LoRA and the task LoRA, at 8 steps with `BasicGuider`. Tested with Character Swap v1.1; Head Swap and the other guide-latent LoRAs load the same way. Keep the image size a multiple of 32 (the workflow resizes to a 2048 px long side). ## Common setup Place LoRAs in `ComfyUI/models/loras/`. The included workflows list their required models and custom nodes inside the graph. Keep the output aspect ratio matched to the source clip and use an H3-supported frame count (`17n + 5`: 5, 22, 39, 56, 73, 90, 107, 124, ...). The LMS, Style Transfer, VFX Edit and Head Swap models use an aligned source-video guide through `MiniMaxH3AddGuide` at `frame_idx = 0`. This is different from a native image/video reference: the guide shares the generated video's spatial and temporal grid. ## License and credits Licenses are specified by asset. The separate [Character Swap license](LICENSE_CHARACTER_SWAP.md) and [Motion Transfer license](LICENSE_MOTION_TRANSFER.md) apply only to the Character Swap and Motion Transfer LoRA weight files listed in [LICENSES.md](LICENSES.md); they do not apply to the other LoRAs, workflows, docs, or examples. Use of these LoRAs, including commercial use in videos and films, requires no prior permission. Organizations with over US $1,000,000 in annual gross revenue must credit its origin for commercial use. Offering the LoRA as a hosted API, SaaS, or other remote model service is not allowed. See the full [Character Swap](LICENSE_CHARACTER_SWAP.md) and [Motion Transfer](LICENSE_MOTION_TRANSFER.md) licenses and check each other asset's applicable terms before use. Base model credit to [MiniMaxAI](https://huggingface.co/MiniMaxAI/MiniMax-H3), ComfyUI-ready weights to [Comfy-Org](https://huggingface.co/Comfy-Org/MiniMax-H3), and guide-latent training support to [ostris/ai-toolkit](https://github.com/ostris/ai-toolkit) and [ComfyUI-AIToolkit-MiniMaxH3](https://github.com/ostris/ComfyUI-AIToolkit-MiniMaxH3).