MiniMax H3 Locally Run Issues: Seeking Config for 12GB VRAM
Inugamix · reddit · 2026-08-30
User attempts to run the INT4 quantized MiniMax H3 model locally in ComfyUI for video generation but encounters severe visual artifacts, static noise, motion smearing, or blockiness. Hardware specs include an NVIDIA RTX 3060 (12GB VRAM) and 32GB system RAM, utilizing the default Image to Video template. Previous attempts with YouTube workflows failed to load, and using the Turbo LoRA did not resolve the issues. The user seeks guidance on recommended sampler setups or working node graphs suitable for 12GB cards.
More from Apps
- Dia AI browser officially launches on Windows — charliedeets · 2026-10-16
- Gemini Canvas workflow: turn a photo with measurements into an interactive 3D CAD studio exporting STLs — GeminiApp · 2026-10-16
- Self Labs launches Agent Vault to let AI agents log in and pay with credit cards — EXM7777 · 2026-10-16
- Teachers say students love interactive presentations built with Claude Opus 5.5 — S_Conradi · 2026-10-16
- Firecrawl raises $75M Series B, launches Fireround funding intel in ChatGPT and Claude — mcraddock · 2026-10-16
- Notion launches 150+ MCP connections to pipe HubSpot, Stripe data into agents — ivanhzhao · 2026-10-16