Making a music video with Minimax H3: a full-reference video prompt workflow
Select_Bowler3099 · reddit · 2026-09-08
A Redditor shares the complete workflow behind their first music video made with Minimax H3: R2V + a 4-step LoRA at 0.6 MP, upscaled via SeedVR2.
The centerpiece is a "video prompt engineer" system prompt for full-reference models:
- Fixed six-section output: subjectdefinitions, summary, retentionanalysis, detaileddescription, overallsoundscape, nondiegeticmusic.
- Consistent <Subject N> / <Picture N> / <Video N> / <Audio N> labels referencing people, keyframes, videos, and audio tracks across all sections.
- summary must start with a task-type prefix like [reference generation] or [audio reference].
- Strict faithfulness rules: no added/omitted visual elements or actions, dialogue kept in <d> tags in the original language.
The template is transferable to any video model supporting full-reference mode.
Related event: Open-source Minimax H3 video workflows gain traction(2 posts)→
More from Multimodal
- TischEins, a large Midjourney SREF library, relaunches with new design after security outage — tisch_eins · 2026-10-26
- MOSS-Image-SFT 4-step distillation tops 16.5K downloads on Hugging Face — _akhaliq · 2026-10-26
- Claude Code can now edit videos: a 10-prompt workflow from raw footage to final cut — TawohAwa · 2026-10-26
- One-sentence 3D worlds turn cinematic with CineMap and TapNow — gizakdag · 2026-10-26
- No Camera Needed: Creator Builds an Entire World in CineMap and Directs It Into Cinema — gizakdag · 2026-10-26
- Kling vs. Seedance: the Chinese AI video model rivalry — DavidmComfort · 2026-10-26