Running MiniMax H3 locally: 0.2MP preview looks good, 1MP render takes 8 hours and drifts
ok_n_na · reddit · 2026-11-09
A user running MiniMax H3 locally via ComfyUI's reference-to-video template reports that testing at 0.2 megapixels gives good 14-second results (i9 14th gen, 64GB RAM, RTX 3060 12GB), but raising to 1 megapixel produces inconsistent outputs and takes 8 hours per render. They're asking for a workflow that keeps low-res and high-res results consistent. The same prompt on Pika also "hallucinates" a different result.
More from Multimodal
- The strange beauty of GAN layers, visualized layer by layer — Merzmensch · 2026-11-09
- Solo creator releases AVIDUS ODYSSEY, a 96-minute AI-assisted space opera animated film — Gunkavoider · 2026-11-09
- Trama: a minimalist open-source orchestrator for multi-AI automations on ComfyUI — Cold-Voice2737 · 2026-11-09
- FoundYou: a single 52M-param SAM 2-based model unifies personalized segmentation and retrieval — rsasaki0109 · 2026-11-09
- GLITCH Studio seeks beta testers for all-in-one local AI filmmaking app built on ComfyUI — Upstairs-Giraffe-189 · 2026-11-09
- Local image app ZPix adds MOSS-Image SFT, SDXL and FLUX support — SamuelTallet · 2026-11-09