Generative AI Google Brain

audio-speech-getting-started-get-started-with-gemini-tts-voices

3 min read

Get started with Gemini-TTS voices using Text-to-Speech

Source notebook

Repo path: audio/speech/getting-started/get_started_with_gemini_tts_voices.ipynb · Open on GitHub · intro

Synthesizes Gemini-TTS speech with Cloud Text-to-Speech and Agent Platform APIs.

Summary

This notebook teaches how to use Gemini-TTS voices for single-speaker, multi-speaker, and streamed text-to-speech synthesis. It demonstrates setup, authentication, voice and locale selection, prompt-driven style control, expressive tags, dialogue synthesis, safety filter relaxation, and Agent Platform generation. The workflow compares Cloud Text-to-Speech API and Agent Platform API paths and plays or writes generated audio.

Key code patterns

Initialize Text-to-Speech client

API_ENDPOINT = "texttospeech.googleapis.com"
client = texttospeech.TextToSpeechClient(
    client_options=ClientOptions(api_endpoint=API_ENDPOINT)
)

Creates a Cloud Text-to-Speech client, optionally using a regional endpoint.

Prompted single-speaker synthesis

voice = texttospeech.VoiceSelectionParams(
    name=VOICE, language_code=LANGUAGE_CODE, model_name=MODEL
)
response = client.synthesize_speech(
    input=texttospeech.SynthesisInput(text=TEXT, prompt=PROMPT),
    voice=voice,
    audio_config=texttospeech.AudioConfig(audio_encoding=texttospeech.AudioEncoding.MP3),
)

Shows how prompts steer emotion, tone, pace, and delivery for Gemini-TTS.

Multi-speaker dialogue

multi_speaker_voice_config = texttospeech.MultiSpeakerVoiceConfig(
    speaker_voice_configs=[
        texttospeech.MultispeakerPrebuiltVoice(speaker_alias="Zizu", speaker_id="Fenrir"),
        texttospeech.MultispeakerPrebuiltVoice(speaker_alias="Gary", speaker_id="Orus"),
    ]
)

Maps speaker aliases to prebuilt voices for dialogue synthesis.

Streaming synthesis

def request_generator():
    yield config_request
    yield texttospeech.StreamingSynthesizeRequest(
        input=texttospeech.StreamingSynthesisInput(text=TEXT, prompt=PROMPT)
    )
streaming_responses = client.streaming_synthesize(request_generator())

Streams audio chunks back as soon as they are ready for real-time use cases.

Agent Platform TTS

client = genai.Client(enterprise=True, project=PROJECT_ID, location=LOCATION)
response = client.models.generate_content(
    model="gemini-2.5-flash-tts",
    contents=TEXT,
    config=types.GenerateContentConfig(speech_config=types.SpeechConfig(...)),
)

Uses Google GenAI SDK enterprise mode to synthesize speech through Agent Platform.

Models & APIs used

When to use this

Use this pattern when building prompt-steerable speech synthesis with Gemini-TTS voices, including dialogue or low-latency audio streaming.

Gotchas & caveats

  • google-cloud-texttospeech version 2.31.0 or newer is required for Gemini-TTS fields.
  • A Google Cloud project is required and the Text-to-Speech API must be enabled.
  • The notebook sets ADC quota project and runs application-default login.
  • Regional availability should be checked in Gemini-TTS documentation.
  • Agent Platform API returns PCM 16-bit 24k audio without WAV headers, so client-side conversion may be needed.
  • Cloud Text-to-Speech supports bidirectional streaming, while Agent Platform supports single request and multiple responses.
  • relax_safety_filters is only enabled for accounts with monthly invoiced billing.
  • ffmpeg installation is handled separately for Linux and macOS.

Best practices

  • Choose Cloud Text-to-Speech API when specific output encodings or bidirectional streaming are needed.
  • Choose Agent Platform API when already using Gemini-TTS from AI Studio or other Agent Platform models.
  • Use natural-language prompts to steer style, accent, pace, tone, and emotional expression.
  • Use multi_speaker_voice_config to assign speaker aliases to specific voices.
  • Stream audio chunks immediately to the frontend in real-time applications.
  • Check available voices, locales, and regional availability in the Gemini-TTS documentation.

Graph View

Backlinks