manifest.json Reference
The manifest is the engine's contract with Studio. These fields drive discovery, capabilities, behavior, and distribution.
At a glance
- Identity:
engine_id,entry_class, versions. - Capabilities, languages, and resource profile.
- Behavior metadata (chunk limit, progress).
- A
distributionblock for install/update.
Core fields
| Field | Purpose |
|---|---|
engine_id | Stable id; also the tts_<id> folder name. |
display_name | Shown in the UI and engine browser. |
engine_version / min_app_version | Semver; gates updates. |
entry_class | Path to the engine implementation. |
resource_profile | GPU/VRAM/CPU needs for scheduling. |
supported_languages / supported_voice_asset_types | What it can do. |
requires_network | Cloud engines disclose off-machine calls. |
Distribution block
Engines install from GitHub by git clone and update by git pull. The distribution block carries the git_url, project, discovery topic, and an optional pin_ref. See Installing & Updating Engines.
Behavior metadata
Behavior fields (chunk limits, progress patterns, per-voice settings) let core behavior come from data. See Behavior Metadata.