A 30s opera short storyboard shows the prompt craft behind one-take Seedance 2.5 films

CurieuxExplorer · x · 2026-09-25

• Creator liluocheng13 shared the storyboard for Act 2 "Touch": ChatGPT for images, Seedance 2.5 for a 30s 1080p clip, with precise image placement timestamps, camera moves (85mm slow push-in, 50mm low tracking) and continuity rules (the knife stays on stage, one bound male lead).
• The core constraint is audio: all non-diegetic music is banned; the only score is elderly musicians performing on-screen backstage, with emotion coming from performance, live accompaniment, ambience and final silence.
• It's a replicable template for AI narrative filmmaking: encode the no-score rule into prompts and use storyboards to lock continuity and cinematography.

Original post →

More from Multimodal

• Asked an open AI model to not copy Apple's foldable — it invented a new device — nikola_mr64990 · 2026-09-25
• One-click narration pipeline: Claude writes shot-aware script, ElevenLabs MCP voices it — nikola_mr64990 · 2026-09-25
• Claude imagines a 2076 world where AI wiped out humanity, generating the whole video itself — Hesamation · 2026-09-25
• Krea2 spits out weird colors on ComfyUI even after re-downloading model, VAE and encoders — fernando782 · 2026-09-25
• Talking to an AI Sorting Hat: Vivix A1's Real-Time Avatar Refuses to Budge — azed_ai · 2026-09-25
• kohya-ss quietly released Anima LLLite model, but it only changes facial expressions — Internal_Answer_6866 · 2026-09-25

Multimodal channel →