Automatic Cartoon Video Lab
Universal script in → scene staging + action planning + speaker turns → four distinct characters → WebM export with turn-by-turn dialogue. Uses existing APIs; no Script Writer changes.
1. Script input
2. Master G character library
Character reference sheets are stored in the project. Vision API inspects the five permanent reference sheets; Master G validates every selected pose. No manual per-episode image uploads.
3. Automatic scene timeline 0 scenes · 0 sec
4. Preview and export
Universal scene director: actions are inferred from each character's script action, staging is based on all characters present, and dialogue is split into speaker turns. Export embeds Kalpa audio per turn. Character voices use the existing TTS endpoint with per-character playback-rate variation; this is a prototype distinction, not separate cloned voices. Procedural animation supports common actions and props, but arbitrary complex actions need additional rig/action modules. WebM only; music/SFX and MP4 are not included.
- Scene count comes from the script; no fixed three-scene limit.
- Each scene keeps its intended duration; when a target duration is set, durations are adjusted proportionally to match it.
- Master G validates character IDs; a universal local action director maps common script verbs to motion and staging.
- Kilo free model is optional for shot/pose selection; local fallback keeps the app usable if the free model is unavailable.