Google TTS voices and usage

Call Google TTS with POST /v1/audio/speech. Send a configured model, string input, and top-level voice. The gateway converts the request and processes the audio.

Model catalog

A Google catalog snapshot checked on October 15, 2026 lists five HTTP TTS models, 70 distinct preset voice IDs, and 230 model-voice associations. The official preset page lists 30 voices; the 3.8 series includes 40 more in this snapshot.

Model	Voices in snapshot	List
`gemini-3.8-flash-tts`	70	30 shared voices + 40 new voices
`gemini-3.8-flash-lite-tts`	70	Same snapshot list as 3.8 Flash
`gemini-3.1-flash-tts-preview`	30	Shared voices only
`gemini-2.5-pro-preview-tts`	30	Shared voices only
`gemini-2.5-flash-preview-tts`	30	Present in snapshot; absent from current official support table

Model availability depends on your Tikway account and the current model catalog. The models/ prefix belongs to Google’s catalog; use a configured Tikway ID or alias such as google/gemini-3.8-flash-tts.

The request schema accepts a voice name or ID instead of restricting it to this snapshot. Gemini 3.8 can also use available voices from a Google project. Check model and account access before using an ID.

How to read the voice lists

voice is an exact, case-sensitive name such as Kore, Zephyr, or Arlo.
Voice character, pitch, accent, and suggested uses help you choose a voice. They are not parameters of this endpoint.
The sample language in the snapshot is en-US; this does not mean synthesis is English-only. Language support varies by model.
Click a voice ID to hear its provider sample.

Shared preset voices · 30

These voices appear with all five models in the snapshot.

`voice` (play sample)	Voice character	Pitch
`Zephyr`	Female · bright	high-mid
`Puck`	Male · cheerful	mid
`Charon`	Male · narrative	low-mid
`Kore`	Female · firm	mid
`Fenrir`	Male · passionate	low-mid
`Leda`	Female · youthful	high-mid
`Orus`	Male · firm	low-mid
`Aoede`	Female · airy	mid
`Callirrhoe`	Female · easygoing	mid
`Autonoe`	Female · bright	mid
`Enceladus`	Male · breathy	low-mid
`Iapetus`	Male · clear	low-mid
`Umbriel`	Male · easygoing	low-mid
`Algieba`	Male · smooth	low-mid
`Despina`	Female · smooth	mid
`Erinome`	Female · clear	mid
`Algenib`	Male · raspy	low-mid
`Rasalgethi`	Male · narrative	mid
`Laomedeia`	Female · cheerful	high-mid
`Achernar`	Female · soft	high-mid
`Alnilam`	Male · firm	low-mid
`Schedar`	Male · steady	low-mid
`Gacrux`	Female · mature	mid
`Pulcherrima`	Female · direct	mid
`Achird`	Male · friendly	low-mid
`Zubenelgenubi`	Male · casual	low-mid
`Vindemiatrix`	Female · mild	mid
`Sadachbia`	Male · lively	low-mid
`Sadaltager`	Male · scholarly	mid
`Sulafat`	Female · warm	mid

New Gemini 3.8 voices · 40

These voices appear with 3.8 Flash and Flash-Lite TTS in the snapshot.

Show the 40 new voices

`voice` (play sample)	Voice character	Pitch
`Arlo`	Male · confident and clear	deep bass
`Bodi`	Male · soft and approachable	bass
`Brio`	Male · soft and approachable	low-mid
`Cleo`	Female · warm and vivid	high-mid
`Cruz`	Male · steady and professional	deep bass
`Daro`	Male · confident and clear	bass
`Elio`	Male · calm and reassuring	mid
`Enya`	Female · crisp and professional	mid
`Enzo`	Male · brisk and positive	bass
`Finn`	Male · warm and approachable	bass
`Fola`	Female · clear and friendly	mid
`Gero`	Male · clear and friendly	bass
`Hali`	Female · calm and reassuring	high-mid
`Jett`	Male · warm and empathetic	deep bass
`Jori`	Female · warm and approachable	low-mid
`Kira`	Female · composed and confident	mid
`Knox`	Male · warm conversational	bass
`Koda`	Male · direct and friendly	low-mid
`Lora`	Female · clear and friendly	mid
`Ludo`	Male · confident and clear	bass
`Lumi`	Female · warm and approachable	bass
`Mako`	Male · energetic and vivid	bass
`Milo`	Male · soft and approachable	bass
`Neno`	Male · low and mellow	deep bass
`Nika`	Female · warm and approachable	mid
`Nyla`	Female · crisp and professional	high-mid
`Olin`	Male · light and precise	bass
`Rami`	Male · warm and approachable	bass
`Riko`	Male · confident and clear	deep bass
`Rina`	Female · gentle and relaxed	mid
`Sami`	Female · composed and confident	bass
`Sola`	Female · gentle and relaxed	very high
`Tari`	Female · confident and clear	mid
`Tavi`	Female · soft and approachable	high-mid
`Tova`	Female · resonant and soothing	low-mid
`Varo`	Male · relaxed and casual	bass
`Veda`	Female · warm and approachable	high
`Zali`	Female · warm and empathetic	mid
`Zeno`	Male · composed and confident	high-mid
`Zuri`	Female · natural and clear	mid

Standard request

{
  "model": "google/gemini-3.8-flash-tts",
  "input": "Hello, welcome to speech synthesis.",
  "voice": "Kore",
  "instructions": "Read in a warm, clear style.",
  "response_format": "wav",
  "extra_body": { "language_code": "en-US" }
}

Field	Gateway behavior
`model`、`input`、`voice`	Required. `input` must be a string of at most 4,096 characters.
`voice`	A name or ID string; `{ "id": "Kore" }` is also accepted.
`instructions`	On 3.8, used as style metadata; on 2.5/3.1, used in the older prompt text.
`speed`	Range 0.25–4, default 1. Google approximates this through style instructions, so exact speed is not guaranteed.
`response_format`	Google supports `wav` and `pcm`; default `wav`.
`stream_format`	`audio` returns audio bytes; `sse` returns gateway events.
`extra_body`	Google extensions. Fields may also appear at the body root; root values take precedence.

Provider extensions include language_code, speech_metadata, generation_config, multi_speaker_voice_config, and text contents. Only the verified temperature option is shown under generation_config.

The Google adapter receives the full synthesized result before wrapping it in speech.audio.delta and speech.audio.done. stream_format=sse does not imply low-latency upstream synthesis. WAV has a container header; PCM does not.

Gemini 3.8 and older previews

Gemini 3.8 treats the main text as words to read verbatim. Send style guidance through top-level instructions or speech_metadata.style. The 2.5/3.1 adapter uses the older prompt-text behavior, so instruction effects may differ. Gemini 3.8 can use an existing voice_... or voicekey_... in your Google project. Send it as a voice string or { "id": "voice_..." }. This endpoint uses existing voices; it does not create, clone, or list voices. Those IDs are not documented as supported on 2.5/3.1.

Multiple speakers

The gateway supports up to two speakers. Configure preset voices in extra_body.multi_speaker_voice_config, then assign each turn with extra_body.contents[].parts[].speech_metadata.speaker.

{
  "model": "google/gemini-3.8-flash-tts",
  "input": "A greeting between Joe and Jane.",
  "voice": "Kore",
  "response_format": "wav",
  "extra_body": {
    "multi_speaker_voice_config": {
      "speakerVoiceConfigs": [
        { "speaker": "Joe", "voiceConfig": { "prebuiltVoiceConfig": { "voiceName": "Puck" } } },
        { "speaker": "Jane", "voiceConfig": { "prebuiltVoiceConfig": { "voiceName": "Kore" } } }
      ]
    },
    "contents": [{ "role": "user", "parts": [
      { "text": "Hello, how are you today?", "speech_metadata": { "speaker": "Joe", "style": "cheerful and friendly" } },
      { "text": "I'm well, thanks for asking.", "speech_metadata": { "speaker": "Jane", "style": "calm and relaxed" } }
    ] }]
  }
}

Chat completions

Image generation

Video generation

Embeddings

Transcriptions

Speech synthesis

Decisions

Utilities

Google TTS voices and usage

Model catalog

How to read the voice lists

Shared preset voices · 30

New Gemini 3.8 voices · 40

Standard request

Gemini 3.8 and older previews

Multiple speakers

​Model catalog

​How to read the voice lists

​Shared preset voices · 30

​New Gemini 3.8 voices · 40

​Standard request

​Gemini 3.8 and older previews

​Multiple speakers

Model catalog

How to read the voice lists

Shared preset voices · 30

New Gemini 3.8 voices · 40

Standard request

Gemini 3.8 and older previews

Multiple speakers