Kijai/MiniMax-H3-TAE
Summary
Kijai shares a quickly trained 2D tiny VAE for MiniMax-H3, intended to improve latent previews in ComfyUI. A better alternative trained by madebyollin is also linked.
View Cached Full Text
Cached at: 08/11/26, 08:03 AM
Kijai/MiniMax-H3-TAE · Hugging Face
Source: https://huggingface.co/Kijai/MiniMax-H3-TAE Quickly trained 2D tine VAE for MiniMax-H3.
Not the greatest outcome, still beats latent2rgb for preview purposes.
Update: madebyollin has (as usual) trained a proper one, also works with my node:
Currently only works with theModelPreviewOverride-node in ComfyUI-KJNodes:
Similar Articles
Kijai/MiniMax-H3-experimental
Kijai's MiniMax-H3-experimental is a work-in-progress model featuring w4a8 quantization and int8_convrot VAE support for ComfyUI, released for testing purposes.
Kijai/MiniMax-H3_comfy
A Hugging Face repository hosting MiniMax-H3 models converted for ComfyUI usage, along with a Lightx2v distill LoRA for faster inference.
realrebelai/MiniMax-H3_GGUFs
Hugging Face repository providing GGUF quantizations of MiniMax-H3 models for use with ComfyUI, including directory structure and links to required VAEs.
Comfy-Org/MiniMax-H3
Comfy-Org repackaged MiniMax-H3 model files for ComfyUI, including diffusion models, text encoders, and VAEs, with workflow templates for text-to-video, image-to-video, and reference-to-video generation.
Introducing H3 Max by fal (5 minute read)
H3 Max is a post-trained version of MiniMax H3 optimized for maximum speed, ranking #1 in human preference evaluations for video quality, prompt understanding, and aesthetics while generating videos up to 35x faster than the official endpoint.