Vidu Animated Video Production

Problem

Producing an animated short is not just about generating one image or one clip. A usable deliverable needs a script, style, characters, scenes, shot language, dialogue, audio, and timeline alignment. When each step switches tools manually, assets can drift apart and shot continuity breaks.

Workflow

This skill turns a creative prompt into an executable anime production pipeline, with the main steps as follows:
- Script and style: confirm the synopsis, characters, dialogue, and art direction, such as 3D anime, 2D anime, Chinese classical, cyberpunk, or realistic style.
- Scene assets: generate a scene image, character blocking image, and four-panel angle sheet to guide later camera composition.
- Character assets: check or produce a character three-view sheet, select a voice_id, and generate preview audio to keep character and voice consistent.
- Storyboard generation: write prompts using style / shot size / camera / composition / movement + visual description + image emphasis, keep shots to 2–8 seconds; by default use viduq3 to generate dialogue and sound effects, or use viduq2 with separate TTS.
- Assembly and delivery: use merge_video_audio.py to assemble an MP4 from the timeline config, then output the asset manifest and timeline document.

Notes

It fits teams that need a repeatable anime workflow and creators who want centralized control over characters, scenes, storyboards, and audio. The boundary is strict: all image, video, and speech generation must use Vidu API, not other models. Image and video tasks are asynchronous and require polling the task_id; result URLs expire. TTS returns synchronously. Missing ffmpeg or Vidu credentials can block generation or assembly.