Running MiniMax H3 on an 8GB laptop: full I2V pipeline with configs and failures
DG86 · reddit · 2026-08-31
A user shares a complete hands-on walkthrough of generating a "formal dinner party in a Roman atrium" video locally with ComfyUI on an 8GB RTX 4060 laptop:
- First frame: Z-Image-Turbo (bf16, 8 steps) generated the 864×480 atrium background from a simple prompt in 30-40s; Qwen-Image-Edit (2509 Lightning 4-step LoRA) turned it into a dinner-party scene in 1-2 min, one-shot.
- Video: MiniMax H3 I2V (int8) with the lightx2v turbo LoRA at 6 steps (slightly better detail/audio than the default 4), shift 12/3; roughly 2 min wall-clock per second of video.
- Pitfalls: first attempt had frozen figures, smeared motion, a character clipping through the table, and an unidentified narrating voice in the audio—clearly a failure; later attempts iterated on the setup.
More from Multimodal
- Tokyo's Gemini AI Photo Booth turns visitors into plush toys and dioramas — bradneuberg · 2026-10-16
- Orchestrating a 100-image portrait of humanity for aliens with ChatGPT, Argon and p5.brush — dx8xb · 2026-10-16
- Indie Creator Builds an AI-Native Animated Show Solo — ComprehensiveLime654 · 2026-10-16
- Ideogram teases 4.5 arrival and open weights: 'Not yet, but soon' — East_Shoe8814 · 2026-10-16
- Opus 5.5 turns heads with surprisingly polished vertical video demo — FinanceYF5 · 2026-10-16
- 'Celestial Ocean Reverie' prompt template blends starry skies with dreamlike seascapes — LudovicCreator · 2026-10-16