Documentation
Providers and voices
Local services, remote APIs and the inputs a voice needs.
- New users
- Operators
Local and API routes
Install local components through Manager and models through the app's catalogue. Configure API providers in Providers & services. Speech providers and language-model providers have different roles; enabling one does not configure the other.
The release catalogue includes OpenAI, Gemini, Azure speech routes, ElevenLabs and additional supported services. A custom endpoint requires its own model and voice configuration. Account access, region and deployed model can affect availability.
Finding a model
Use Models and languages to compare model families. Expand Exact variants for the individual checkpoints: language support, reference requirements and controls can differ within a family. The language filter combines equivalent codes and regional variants under readable names, while keeping distinct languages and dialects separate.
The source links lead to model cards and licences for the weights. API entries link to their providers’ terms. Pandrator’s MIT licence does not replace these terms.
Voice choices
A prebuilt voice is selected by its provider or model ID. Cloning uses a reference recording and, for some models, its transcript. Voice design generates a synthetic voice from a description. RVC converts already-generated audio through a separate voice-conversion model.
A saved library reference does not become compatible with every engine. Check the current model's requirements and audition it with that reference.
Provider inputs
Remote speech services may receive narration text and, when supported, a voice sample. Remote correction and translation services receive text and context. Store credentials in the app or Manager's settings; give agents the provider identity rather than its secret key.