4090 Tested: Ultimate ComfyUI Setup for High-Speed Minimax H3 Video Generation
BigWideBaker · reddit · 2026-08-05
The author shares a ComfyUI configuration guide to significantly boost Minimax H3 video generation speeds on an RTX 4090. Key optimizations include:
- Environment: Using the latest portable ComfyUI, Python 3.13, PyTorch 2.13, and installing KJ nodes directly from GitHub to access the experimental MiniMax H3 Mem Eff Sage Attention Patch.
- Attention Mechanism: Installing Sage Attention via ComfyUI-Sage-EasyInstall.
- Workflow Parameters: Utilizing euler sampler and linearquadratic scheduler at only 15 steps with fp16 accumulation enabled, drastically reducing gen time with little to no quality loss.
- Model: Using the minimaxh3fl2vaprunedint8convrot.safetensors quantized model.
Related event: MiniMax H3 Local Tests: Performance and Optimization Across GPUs(13 posts)→
More from Multimodal
- Tessellation-free geometric queries on displaced surfaces cut caustics variance 10x — ssh4net · 2026-10-16
- Complex multi-link CAD mechanism finally animates, dev celebrates the breakthrough — zetalyrae · 2026-10-16
- OpenMOSS releases MOSS-Image-SFT: Apache 2.0 text-to-image model on Hugging Face — bdsqlsz · 2026-10-16
- Hyperflow open-sourced: 49 sampling steps down to 8, ~3x faster video generation — mhdfaran · 2026-10-16
- Cinematic 3D fantasy character prompt template with salvaged-robot armor — azed_ai · 2026-10-16
- Midjourney vintage woodcut style: --sref 2446102396 beaver with a cleaver — michaelrabone · 2026-10-16