comfyui-workflows/models/minimax/minimax-h3-int8-r2v-javano2608-13/README.md

MiniMax H3 INT8 Reference-to-Video

Summary

MiniMax H3 reference-to-video workflow with INT8, Motion Context, and a dedicated reference LoRA path.

Model Family

• minimax

Status

• imported-only

Source

• Workflow repo: https://huggingface.co/javawock7618/comfy-MiniMax-H3-workflows
• Workflow file: https://huggingface.co/javawock7618/comfy-MiniMax-H3-workflows/resolve/main/MiniMax_int8_R2V-javano2608.13.json?download=true
• Original publisher: javawock7618
• Date imported: 2026-08-12

Developer notes

• This is the R2V branch from the source pack.
• The source README describes it as an INT8 workflow built around:
  • Sage-Attention
  • Spectrum
  • Lightx2v LoRA
  • optional Turbo-LoRA
  • Motion Context
  • latent upscaling
• The source README says the R2V branch expects minimax_h3_ref_lora_rank_256_bf16.safetensors.
• The pack also ships FLF and TTS workflows, but this import is the dedicated reference-video graph.

Our notes

• This is the strongest same-day MiniMax H3 workflow hit because it is actually about reference conditioning and consistency, not just another turbo LoRA.
• It is still a heavy ask for an RTX 3070 / 8 GB, but it is much more likely to teach you something useful than a novelty-only MiniMax drop.
• Best MiniMax workflow to test if you want a real comparison point against LTX 2.3 consistency work.

Required custom nodes

• ComfyUI-Spectrum-MiniMax-H3
• ComfyUI-H3-Motion-Context
• ComfyUI-MiniMaxH3_LatentUpscaler
• ComfyUI-SolAttn_triton
• ComfyUI-GGUF if you swap to GGUF-based variants
• comfyui-easy-use
• comfyui-kjnodes
• comfyui-impact-pack
• comfyui-ollama
• comfyui-videohelpersuite
• WhatDreamsCost-ComfyUI
• comfyui_nvidia_rtx_nodes
• was-node-suite-comfyui

Required model files

• Comfy-Org/MiniMax-H3
• minimax_h3_ref_lora_rank_256_bf16.safetensors from Kijai/MiniMax-H3-experimental
• optional Kijai/MiniMax-H3_comfy
• optional drbaph/MiniMax-H3-Turbo-Lora-ComfyUI

Notes

• This is the best current MiniMax reference-video import in the repo.
• If your goal is faster low-step novelty renders, the turbo paths are easier to demo, but this one is the better consistency-oriented test.