VEDA Sparse Attention Models to get x1.7-x2 boost speed
custom node https://github.com/veda-sparse/Veda-on-ComfyUI. That node also is in Manager
- [minimax_h3_t2va_veda_8nfe_600step_preview_fp8.safetensors](https://huggingface.co/Veda-Sparse/Minimax-H3-T2VA-Veda-8NFE-600Step-Preview/resolve/main/minimax_h3_t2va_veda_8nfe_600step_preview_fp8.safetensors)
- [minimax_h3_r2va_veda_preview_fp8.safetensors](https://huggingface.co/Veda-Sparse/Minimax-H3-R2VA-Veda-Preview/resolve/main/minimax_h3_r2va_veda_preview_fp8.safetensors)
Set Veda Settings like this to get stable results and avoid random flashes. Works with FL2VA and Ref2VA
File save location
ComfyUI/
│ ├───📂 models/
│ │ └─── 📂 veda/
│ │ └─── minimax_h3_t2va_veda_8nfe_600step_preview_fp8.safetensors
│ │ └─── minimax_h3_r2va_veda_preview_fp8.safetensors
All models download from https://huggingface.co/Comfy-Org/MiniMax-H3/tree/main
Heretic Text Encoder int8- https://huggingface.co/ethanfel/Qwen3-VL-32B-Ultra-Heretic-H3-ComfyUI-INT8-ConvRot/tree/main
Heretic Text Encoder nvfp4 - https://huggingface.co/sakamakismile/Qwen3-VL-32B-Heretic-MiniMax-H3-NVFP4/tree/main
My ComfyUI Build - https://huggingface.co/datasets/StefanFalkok/ComfyUI_portable_torch_2.14.0_cu130_cp313_sageattention_triton
My NSFW TG Channel - https://t.me/stefanfalkokmoth
My TG Channel - https://t.me/StefanFalkokAI
My TG Chat - https://t.me/+y4R5JybDZcFjMjFi
Description
Deleted Spectrum nodes because of the turbo lora with better quality in I2V Workflows (with fl2va models)
FAQ
Comments (5)
Can this workflow work with 4gb vram and 16gb of ram?
I'm not sure about that
I approve this workflow. I send you tip now. Thank you.
minimax is a video generator that uses image checkpoints? is it like Wan with models and uses its text encodes etc?
yeah, Minimax supports I2V