Real footage to 1980s OVA anime: full LoRA training workflow for MiniMax H3 reference-to-video
ostrisai · x · 2026-11-07
JulienAIArt shares a complete style-LoRA training workflow for MiniMax H3 Reference-to-Video (real video in as reference), trained locally with AI Toolkit on a Dell tower with an RTX PRO 6000. The dataset has two parts: 21 image pairs (real photo vs. same photo redrawn as OVA anime) to teach the look, and 10 short anime clips paired with themselves as the answer to teach the model to copy motion rather than invent its own. The result keeps poses, outfit and movement while changing only the style.
Training details: 30 hours, 4,000 steps, rank 16. Loss flattened at step 1,000 but samples kept improving until 2,500; by 4,000 it was overtrained. Rendered in ComfyUI. A detailed talk follows at the Dell booth, Adobe MAX, Nov 10-11.
More from coding & agent
- Gemini /teamwork agents caught live: four parallel task logs all running that second — RyanSpunk · 2026-11-13
- Codex CLI 0.168.0 adds model provider picker, side-by-side TUI workspace and animated GIF generation — github-actions[bot] · 2026-11-13
- Free MCP server lets Claude read Altium, Allegro, KiCad and Gerber PCB files directly — woodisruff · 2026-11-13
- a16z Partner Clones Open-Source Game with Fable 5.2 Agent, Which Finds Its Own Horse Animation Asset — FinanceYF5 · 2026-11-13
- OpenAI team shares how it runs and governs fleets of persistent agents internally — romainhuet · 2026-11-13
- Perplexity's Decider scanned 1B tokens across 7,000 agent sessions, found 14 security fixes — denisyarats · 2026-11-13