MATLOWAI/minimax-h3-fused-turbo-int8-convrot
image-text-to-video model by MATLOWAI
Older snapshot: check the original source for current information.
From the publisher
Model card & documentation
Source previewRead the publisher’s intended use, setup instructions, evaluations and limitations. The original model card is the source of truth.
One 21 GB ComfyUI diffusion-model file that does MiniMax-H3 text/image-to-video and reference-to-video in 4 steps, with a distilled turbo and a motion-smoothing style LoRA already folded into the weights. so a single partition serves both first/last-frame and reference conditioning (xmarre/MiniMax-H3-Pruned-Ref-Delta-Fused-r1024-ComfyUI, ComfyUI conversion of diffusers-modular/MiniMax-H3-Pruned-Ref-Delta-Fused-r1024). Mystic v2.0 at 0.7 (Civitai model 2856467, merged for motion smoothness). (qkvproj, outproj, fc1, fc2) are INT8 with ConvRot (group 256, per-channel) in ComfyUI's native comfyquant layout; everything else stays BF16/F32. No custom loader: UNETLoader, weightdtype default. diffusionmodels/minimaxh3fusedrefdeltar1024turbo8mystic07int8convrot.safetensors -> ComfyUI/models/diffusionmodels/ Hardware. The timings below are from a 96 GB card with everything resident. Smaller cards work too: ComfyUI's model management loads and offloads the transformer, text encoder and VAEs as they are needed, so the same graphs run, just slower, and the 21 GB file is the smallest footprint this recipe has. AMD / ROCm: the recipe was developed on an AMD box; use patientx's ComfyUI…
Read the full model card ↗ · Preview checked 2026-10-31T07:42:32.556Z
Inside the original model card — Document outline
- MiniMax-H3 Fused Turbo (INT8 ConvRot)
- Results and timings
- Don't want the bake? The live-LoRA path is equivalent
- Why merge instead of stacking the LoRAs
- Steps: it says 8, run it at 4
- The recipe that made the previews
- Workflows (workflows/)
- Low VRAM (workflows/lowvram/)
- Motion adapters
- Credits and links
- License
Headings are captured from the source. Links open the publisher’s document, not a locally hosted copy.
Documentation belongs to its respective authors. Reported project/model license: other. A listing is not a grant of reuse or training rights. Confirm the document’s own terms at the source.
Model overview
MATLOWAI/minimax-h3-fused-turbo-int8-convrot is a image-text-to-video model repository published by MATLOWAI on Hugging Face. The source reports the diffusion-single-file library.
This page summarizes Hub metadata. For intended use, training data, evaluation results and limitations, consult the original model card.
Model facts
- Task
- image-text-to-video
- Library
- diffusion-single-file
- Recent downloads (30 days)
- 452,396
- Cumulative likes
- 185
- Hugging Face trending score
- 52
- Reported safetensors parameters
- Not reported
- Architecture
- Not reported
- License
- other
- Access
- Not gated by Hugging Face
- Created
- 2026-08-29T22:04:12.000Z
- Last modified
- 2026-08-30T01:25:46.000Z
Compatibility and lineage
Language coverage was not reported.
Reported base models: xmarre/MiniMax-H3-Pruned-Ref-Delta-Fused-r1024-ComfyUI, diffusers-modular/MiniMax-H3-Pruned-Ref-Delta-Fused-r1024, Comfy-Org/MiniMax-H3, MiniMaxAI/MiniMax-H3
Parameter count is not a RAM/VRAM requirement. Check precision, quantization, context length and runtime compatibility in the model card; no hardware or API-cost claim is inferred here.
Use and evaluate
Check license terms and access requirements first. Review the model card’s documented loading instructions, evaluations and safety limitations. Benchmark scores are not imported by this directory, and popularity is not an accuracy ranking.
No model weights are downloaded or executed by AltAPIs. Never enable remote model code without reviewing it.
Source and freshness
Source: Hugging Face Hub. Metadata observed 2026-09-15T06:34:02.617Z. Daily imports are snapshots, not real-time monitoring.
Popularity and source listings do not establish security, suitability, licensing rights or benchmark performance.
Related models
pablodawson/MiniMax-H3-360-Orbit-LoRA