Triple MiniMax H3 Speed in ComfyUI with CUDA 13 and INT8
AI-imagine · reddit · 2026-08-04
This tutorial details how to triple the video generation speed of the MiniMax H3 model in ComfyUI by enabling true CUDA 13 (cu130) and INT8 support.
Core Optimization Steps:
- Update PyTorch & CUDA: Upgrade ComfyUI's embedded Python environment to a PyTorch version supporting cu130.
- Reinstall Attention Libraries: Uninstall old sageattention and flash-attn, then install the SageAttention 2.2.0 wheel built for cu130.
Performance Comparisons & Node Recommendations:
- Base Speedup: Before properly enabling INT8, a 5-second video took 12-13 minutes; after the fix, it takes just 4 minutes.
- EasyCache Node: Adding this node slashes generation time for a 15-second 480p video from 21 minutes to 11 minutes with almost unnoticeable quality loss and identical audio.
- Further Compression: Combining EasyCache with 13 steps reduces time to 8 minutes, retaining 90% of motion coherence with only minor voice detail loss—perfect for quick previews.
The author notes that fully optimized, MiniMax H3 is miles ahead of models like Wan2.1 in prompt following, audio matching, and reference image control.
Related event: MiniMax H3 Local Tests: Performance and Optimization Across GPUs(13 posts)→
More from Multimodal
- Seedance 2.5 workflow: character reference sheets + structured prompt yields a 30s single-take fake tourist video — techhalla · 2026-10-16
- AI couple moves into a castle ruin: full Magnific + Seedance 2.5 prompt thread behind the viral clip — techhalla · 2026-10-16
- FLUX.3 Klein shines at consistent high-res photo editing on RTX 5090: 3000×4000 in ~230s — herbertseabra · 2026-10-16
- Smoky Velvet Noir: a reusable image-gen prompt template with three moody examples — LudovicCreator · 2026-10-16
- VideoGen-Agent: An 8B Multimodal Agent Turns Video Generation Into a Tool-Use Loop — jiqizhixin · 2026-10-16
- Redditor builds impressive seaplane sim project with Sonnet 5.5 + Blender — Wise_Shame_2853 · 2026-10-16