Storyboard grids tame MiniMax reference-to-video for complex music video scenes
Cheap_Credit_3957 · reddit · 2026-10-26
A music video creator shares a workflow for complex scenes that MiniMax reference-to-video couldn't handle with prompts or first/last frames alone: generate a 2x2 storyboard grid with GPT image 2 (panels = start, progression, end), feed it with character/location references into MiniMax H3 Singularity, and structure prompts so each shot "hard cuts, exactly like panel N." Technical details: two-pass sampling (25 steps, then an 8-step 768p turbo LoRA at 4 steps, 0.35 denoise with latent upscale); stage-performance shots still use normal reference-to-video. Full ComfyUI video builder and two-pass workflow are on GitHub. The author suspects the separate character/location references may not even be needed.
More from coding & agent
- Recreating Kurzgesagt's Entire Channel in 10 Minutes with a Claude Prompt — zack_overflow · 2026-10-26
- DeepSeek V4.1 Flash with OpenCode Go: cheap unlimited coding agent with Opus validating results — haider1 · 2026-10-26
- No coding agent vendor has shipped DeepSeek's open-sourced DeepInfer yet — pingzawa162 · 2026-10-26
- Agent payments are already live via x402 — but what happens when a transaction stalls mid-payment? — SpreadUsual4084 · 2026-10-26
- Connecting Copilot Studio to SAP S/4HANA via MCP with per-user identity — ColdImportance2484 · 2026-10-26
- Base44 fine-tuned Qwen 2B in half a day for prompt autocomplete serving millions — tristanbob · 2026-10-26