Anime MV with MiniMax H3: a full workflow from Suno song to sakuga-style animation
Gorwald · reddit · 2026-10-05
A creator shares a complete workflow for making anime-style animation with MiniMax H3, after a month of struggling with a too-3D/cel-shaded look until a YouTube tutorial on Seedance 2 prompts (DBZ-style) pointed the way.
Workflow:
- Make a fast-paced song in Suno and imagine a sakuga-style anime opening;
- Have ChatGPT design characters and scenes with the directive "2D semi-flat colors, clean design, not super busy";
- Use ChatGPT to map the prompt structure onto the song structure, scene by scene;
- In ComfyUI, use foxdit's minimaxH3SEEDHUNTERLatentv21 workflow + Dasiwa hybrid V3 checkpoint + Dasiwa RTX Upscaler node (impressively good).
Pitfalls: the model adds unwanted fast character turn-backs, and once you tell ChatGPT "this zoom looks cool," it adds multi-zoom helmet shots every time — prompt habits stick.
More from Multimodal
- ComfyUI's janky outputs make surprisingly great surreal horror — p0lar0id · 2026-10-05
- LTX releases Alpha Matting IC-LoRA: VFX-grade transparent matting in 20 seconds — multimodalart · 2026-10-05
- Prompt template: Cubic Dimensional Shift for interlocking-cube image effects — LudovicCreator · 2026-10-05
- Plenio turns YuE2 into a DAW: an AI music workstation inside ComfyUI — Vivid_Promise1700 · 2026-10-05
- Cute "Claude hits the beat" animation made with Fable 5.5 — ring_hyacinth · 2026-10-05
- Realistic consistent AI characters on $10-20/month: Flux dev + LoRA on RunPod — SomeCartographer4601 · 2026-10-05