MiniMax H3 anime testing: SLA attention plus 4-step LoRA runs on a 3080 Ti laptop
Working-Distance-901 · reddit · 2026-09-06
Reddit user Working-Distance-901 tested MiniMax H3 anime generation using the Ref2VA pipeline with SLA attention and a 4-step LoRA for fast inference. On a 3080 Ti laptop with 16GB VRAM, 5-second videos take 250s and 10-second ones 600s. Outputs are upscaled to 1.0 MP via RTX Video Super Resolution, with music generated by MiniMax Music 3.
More from Multimodal
- AI video now pulls off orbiting, diving camera moves around a speeding bicycle — umesh_ai · 2026-10-23
- AI art showcase: gothic winged warrior "Azrael-42" with porcelain-crack armor — LudovicCreator · 2026-10-23
- Moody AI-reimagined Harry Potter portraits showcase stunning cinematic generation — azed_ai · 2026-10-23
- Claude Code + ComfyUI MCP workflow for client work: paid models or open weights? — newaccount47 · 2026-10-23
- AI-generated character sheet wows: melancholy traveling cat 'Finch' fully spec'd — azed_ai · 2026-10-23
- Viral prompt template turns AI image models into children's book character-sheet generators — azed_ai · 2026-10-23