Skip to main content

Gemini TTS

Gemini TTS uses Neiroha's Google Gemini TTS adapter and calls the native text-to-speech capability of the Gemini API. It is useful for small tests and style experiments with a Google AI Studio API key.

Official Limits

Google marks Gemini TTS as Preview. Official rate-limit pages currently list gemini-2.5-flash-preview-tts free-tier limits as 3 RPM / 10,000 TPM / 15 RPD. Quotas, model names, and availability can change; use the AI Studio console as the source of truth.

Official pages:

Provider Fields

FieldRecommended Value
Adapter TypeGoogle Gemini TTS
NameGoogle AI Studio
Base URLhttps://generativelanguage.googleapis.com
API KeyGoogle AI Studio API key
Default Modelgemini-2.5-flash-preview-tts by default

After saving:

  1. Click Fetch All.
  2. Neiroha shows built-in Gemini TTS models and official preset voices.
  3. Enable the provider.
  4. Click Health Check.
  5. Create a preset-voice character in Voice Bank.

Voices

Gemini TTS official docs list 30 preset voices. Neiroha shows them as a fixed list, including:

Zephyr, Puck, Charon, Kore, Fenrir, Leda, Orus, Aoede,
Callirrhoe, Autonoe, Enceladus, Iapetus, Umbriel, Algieba,
Despina, Erinome, Algenib, Rasalgethi, Laomedeia, Achernar,
Alnilam, Schedar, Gacrux, Pulcherrima, Achird, Zubenelgenubi,
Vindemiatrix, Sadachbia, Sadaltager, Sulafat

For the first test, Kore or Puck with short text is a good starting point.

Character Settings

GoalCharacter Setting
Normal readingUse preset voice mode and select a Gemini preset voice.
Style controlPut director-style hints such as "gentle, low voice, news style" in the character voice instruction.
Voice cloningNot supported; the Gemini TTS adapter rejects reference-audio cloning.

Gemini TTS has no dedicated speed field. Neiroha converts non-1.0 speed choices into natural-language style prompts.

Free Quota Usage

Gemini free-tier request counts are suitable for Quick TTS and short Dialogue tests, not complete novels.

Recommended provider limits:

FieldValue
Max concurrency1
RPM3 or lower
TPM10000 or lower
RPD15 or the value shown in your console

If you hit 429 RESOURCE_EXHAUSTED, pause requests, then lower concurrency and batch size.