Skip to content

gemini_config.md

Latest commit

 

History

History
53 lines (38 loc) · 1.65 KB

gemini_config.md

File metadata and controls

53 lines (38 loc) · 1.65 KB

Gemini TTS Configuration

Gemini TTS uses a dedicated config file instead of Azure's .env or Polly's AWS config path. Keeping Gemini auth and region settings in their own file makes provider-specific failures much easier to trace.

Recommended Path

  1. Create a local .gemini.env file in INI format.
  2. Create or choose a Google Cloud service account JSON file with access to Cloud Text-to-Speech and the aiplatform.endpoints.predict permission.
  3. Open Tools > Settings.
  4. Set TTS Provider to Gemini TTS.
  5. Point Gemini Config File at the dedicated config file.
  6. Choose a Gemini model, language code, voice, and optional style prompt.
  7. Use Test Gemini TTS.

Expected .gemini.env Format

[GEMINI]
project_id = YOUR_GOOGLE_CLOUD_PROJECT_ID
service_account_json = C:\path\to\service-account.json
region = global

project_id and service_account_json are required. region defaults to global if omitted.

Supported Models

The app currently exposes these Gemini-TTS model choices:

  • gemini-2.5-flash-tts
  • gemini-2.5-pro-tts

Prompt And Input Limits

Based on the current Google Cloud Gemini-TTS docs:

  • the text field can be up to 4,000 bytes
  • the prompt field can be up to 4,000 bytes
  • text + prompt can be up to 8,000 bytes total

If those limits are exceeded, the app raises a clear Gemini-specific error instead of sending a bad request.

Notes

  • Gemini TTS is prompt-controlled and does not use SSML in this app path.
  • Gemini-TTS supports style prompts, single-speaker output, and multi-speaker output at the provider capability level.
  • Region/model availability depends on the configured Google Cloud region.