Training an original-character LoRA for MiniMax H3: full workflow and pitfalls
DeliciousGorilla · reddit · 2026-10-21
A detailed walkthrough of training an original-character LoRA for MiniMax H3 video: dataset from 8 Z-Image stills + 14 Seedance clips + ElevenLabs voice, 41-minute H200 training in Fizgig, RTX 3090 rendering with DLSS 5, plus key lessons (LoRA 0.8 sweet spot, don't trust trainer previews, RefMod lighting leakage) and shareable workflow files.
More from coding & agent
- Voice agent developers wrestle with the hardest edge case: knowing when to hang up — Kindly-Duty272 · 2026-10-21
- OpenCode hacked, eMudra used private key of 4, PHP hit by supply-chain chaos — burny_tech · 2026-10-21
- PHP ships emergency fixes for two 9.2/10 RCE vulnerabilities — MickeySteamboat · 2026-10-21
- Anthropic releases Claude Sonnet 5.5: Terminal-Bench jumps from 10.3% to 70.6%, tasks up to 30% cheaper — Aiden_Tech_Ai · 2026-10-21
- Indie hacker ships 8 email product features in one drop with Opus, shares inbox open-rate data — marclou · 2026-10-21
- Solo dev runs 6,000+ unattended news episodes in 19 months — eval loops, not models — jwdNyc · 2026-10-21