Google AI Studio
Use Google's Gemini speech API directly with an API key from Google AI Studio.
- Open Settings → Admin → Providers → Add provider.
- Choose Google AI Studio and enter your Google API key.
- Choose Gemini 3.8 Flash TTS (the default) or Gemini 3.8 Flash-Lite TTS.
- Save, then use Check to verify access and generate a short worker sample.
- In Model & Voice, choose the provider and a voice. New setups use Zephyr.
Both models offer 30 prebuilt voices. Custom voice ID accepts additional library or custom voice IDs supplied by Google. Create/manage those voices in AI Studio; OpenReader does not upload recordings or create voices.
Voice instructions controls delivery, such as Calm, warm narration.
Instructions are sent separately from the passage using Google's speech
metadata. Numeric generation speed is unavailable; audio player speed still
works with generated audio.
The worker sends bounded text pieces to Google's Interactions API and converts the returned WAV audio for the existing playback, alignment, cache and export paths. The key remains encrypted in OpenReader and reaches only Google's fixed endpoint through the credential broker. No additional environment variable is needed. Request failures never switch to a different model.
The generic gemini-flash-latest text model is not a speech model. Google AI
Studio keys also differ from Google Cloud Text-to-Speech service credentials.
Replacing the removed provider
DeepInfra is no longer offered. The upgrade removes its saved provider entries, encrypted keys and any installation default pointing at one, following the earlier Speech SDK removal. Choose a remaining provider or add Google AI Studio with a new Google key. Stored document audio is not deleted or relabeled.
API contract: Google speech generation.