MiniMax H3 60-second seamless video workflow optimized for 12GB GPUs
vortis23 · reddit · 2026-09-08
A Reddit author built a VRAM-optimized ComfyUI workflow around j955229's H3 Motion Director node, enabling 60-second seamless multi-shot videos (text-to-video, image-to-video, reference-to-video) on 12GB GPUs at 35 min per minute of video in Turbo mode (8-step distilled LoRA) or 20-step full-precision rendering.
Key trick: a Unified Boolean Toggle syncs model weights and scheduler steps via the native strengthmodel parameter, avoiding the notorious COMFYMATCHTYPEV3 wildcard validation error. Requires ComfyUI v0.34+ plus four custom nodes (Motion Director, kjnodes for SageAttention, diffaid-patches, h3-multishot) and int8-pruned MiniMax H3 UNET, Qwen3-VL-32B text encoder, turbo LoRA, and VAE weights. Workflow JSON is downloadable from CivitAI (Mega mirror provided).
More from coding & agent
- Dev hacks reMarkable e-ink tablet to let Opus draw and animate — Stefania_druga · 2026-10-26
- Godot games are trivially decompilable — compile C# to native DLL for real protection — mflux · 2026-10-26
- Techhalla shares full 2D-to-3D workflow with Opus 5.5 and Kling 4.0 prompts — techhalla · 2026-10-26
- Jason Furman: Building Econvitals pre-AI would need ~50 people, 3 years, $20M — HarrySurden · 2026-10-26
- Dev dictates to Claude Code for 4+ hours while riding indoor bike in aero position — jacob_posel · 2026-10-26
- Proxmox and LXCs are underrated in the age of AI agents — basedjensen · 2026-10-26