Practicing H3 video editing: precise timestamped prompts turn it into T2V
SulpharTriangle · reddit · 2026-10-10
Hands-on findings with MiniMax H3's video edit capability: to make the model follow the source video, you need to spell out actions, motions and camera moves as exactly as possible with exact timestamps — but then the result feels like text-to-video and lands closer to reference generation than true video editing. The author notes akatz-ai's Character Swap LoRA is genuinely impressive, but only fits character swaps, not edits requiring video continuation. For longer single-take sources, continuation is required, and latent handovers like Motion Context practically override video edit instructions; the Add Guide node has the same issue. The author asks the community for better workflows.
More from coding & agent
- Debugging 9K lines of 100% AI-generated code: standard tactics failed to find root cause — blaizedsouza · 2026-10-11
- ManimGX open-sources a Rust+wgpu 3D animation engine for agents, 90× faster than ManimCE — Scobleizer · 2026-10-11
- Redot Engine welcomes AI-generated PRs — but only if contributors understand their own code — esrtweet · 2026-10-11
- $500 ex-mining BC-250 cluster runs Qwen 35B at 145 tok/s with 256k context — Ok-Breadfruit-3523 · 2026-10-11
- Jev founder's harness guide: coding agents up to 200x faster, 400x cheaper — blaizedsouza · 2026-10-11
- First 24 hours with Hark agent: auto-connects accounts, fixes its own Notion errors, 2FA remains a hurdle — Scobleizer · 2026-10-11