Full AI rap MV pipeline on a single RTX 5080: 214 images, 114 H3 clips, Suno audio
circlenline · reddit · 2026-09-07
A creator shared the complete pipeline for making a professional-looking AI rap music video on a single RTX 5080 (16GB VRAM) + 96GB RAM:
- Images: 214 scene images generated with Krea2 Turbo NVFP4 (curated to 82, 40 used), with LoRAs for identity consistency; Flux Klein 9B FP8 used for editing, masking, and detail fixes via the NKD Klein Tools node pack.
- Video: 114 total clips from MiniMax H3 (37 B-roll + 77 rap-performance generations with lip sync). Ref2V took a character reference, a scene image, the matching audio fragment, and exact lyrics in the prompt. Each 5-second 1MP/32-step generation took 7-8 minutes; 15-second generations proved impractical without a Turbo LoRA, so the song was split into short audio fragments.
- Prompting: a custom ChatGPT GPT loaded with H3 docs, PC specs, the workflow, and character-consistency rules generated per-scene prompts.
- Music: Suno track extended via the free model's "extend" after a 1-minute paid-tier preview, merged in editing. Final cut done manually in After Effects. The full workflow is public on a GitHub Gist.
More from coding & agent
- MCP fault injection study: agent task scores drop 60% under realistic tool failures, scale doesn't help — dair_ai · 2026-10-26
- Codex manager pattern: one orchestrator agent directing three specialized peer agents — testingcatalog · 2026-10-26
- General agents burn ~140M tokens/month on LinkedIn outreach; scripts do it 60x cheaper — ChrisHarpon2 · 2026-10-26
- OpenAI upgrades Decisions API: 400 questions per call, native voice and image support — OpenAIDevs · 2026-10-26
- NLI alone rejects 5/6 new agent memories; a 0.45 similarity gate fixes it all — Alternative_Pin9598 · 2026-10-26
- LangChain ships Managed Deep Agents 0.9, walkthrough from Interrupt London keynote — LangChain · 2026-10-26