VisoTalk

The IR

The document an agent writes, a person edits, the player plays and the exporter renders.

The IR is a JSON document, typed in TypeScript and versioned with SemVer. Its current version is 0.2; everything in 0.x may change, and every change bumps the minor version and ships a migration.

Four ideas shape it:

  1. Semantic time. Visuals are anchored to sentences and words of the narration, never to guessed frame numbers.
  2. One document, many outputs. Interactive playback and video export come from the same document and the same DOM renderer.
  3. Object identity. Objects persist across beats, and changes between beats interpolate by identity.
  4. Production built in. Text-to-speech and timing, captions with separate pronunciation, loudness, media licensing and attribution, and layout QA before rendering.

Quick start

Validate, play and render an example document locally.

Examples

Four documents that ship with the engine.