Custom Models
Connect OpenAI-compatible audio transcription and chat completion endpoints.
Open AI Models → Custom to add transcription or enhancement definitions. These are connections to an existing service, not downloaded models.
Custom Transcription
Add an entry in the Custom Transcription Models section and provide:
| Field | Value |
|---|---|
| Display Name | A name shown in VoiceInk. |
| API Endpoint | The full audio transcription URL, including its path, such as https://your-service.example/v1/audio/transcriptions. |
| API Key | The service's API key. |
| Model Name | The exact model identifier accepted by that service. |
| Multilingual Model | Whether the model supports multiple languages. |
The endpoint must accept OpenAI-compatible audio transcription requests. Custom transcription is batch-based and does not provide real-time partial text.
Custom Enhancement
Add an entry in the Custom Enhancement Models section and provide a display name, Base URL, API key, and model name. Despite the field's name, enter the full chat completion endpoint, such as https://your-service.example/v1/chat/completions, including the endpoint path.
The service must accept OpenAI-compatible chat completion requests. The dialog requires a nonempty API key as well as the other required fields.
Test and Save
Use Test if you want to make a connection request before saving. Testing is optional; saving a definition does not automatically verify that the endpoint, key, or model works.
After saving, choose the custom transcription model or enhancement provider in a Mode. Requests go to the URL you configured, which may be local or hosted.
Edit or Delete
Use the model row's menu to edit or delete an entry. Deleting also removes its associated saved Keychain API key.
Settings backups include custom transcription definitions without their keys. They do not include custom enhancement provider definitions.