List voices

GET https://api.kittenml.com/v1/voices?model=kitten-tts-2-latest

Kitten TTS 2 only

The three KittenTTS 0.8 models have a fixed eight-voice catalog and no voice listing. See feature support by model.

Use this authenticated endpoint instead of hard-coding the voice catalog. It returns all 47 Kitten TTS 2 built-in voices with their title and language, the voice-cloning capability, the active saved voices owned by the caller's organization, and any shared voices that KittenML makes available to every organization.

Query parameters

Parameter	Default	Accepted values
`model`	Kitten TTS 2	`kitten-tts-2-latest`
`limit`	`100`	`1` through `100`. Limits saved voices only, never built-ins or shared voices
`after`	—	The `last_custom_id` of the previous page

Always send model=kitten-tts-2-latest. The model fields in the response repeat the ID you sent. A request that omits model still gets the Kitten TTS 2 catalog, but reports it under the model's earlier ID; see the legacy note in Models. The three KittenTTS 0.8 models have a fixed eight-voice catalog, listed in KittenTTS 0.8 voices.

cURL

curl -L --fail-with-body -sS \
  'https://api.kittenml.com/v1/voices?model=kitten-tts-2-latest' \
  -H "Authorization: Bearer $KITTENML_API_KEY"

Python with httpx

pip install httpx

import os

import httpx

response = httpx.get(
    "https://api.kittenml.com/v1/voices",
    params={"model": "kitten-tts-2-latest"},
    headers={"Authorization": f"Bearer {os.environ['KITTENML_API_KEY']}"},
    follow_redirects=True,
    timeout=30,
)
response.raise_for_status()
catalog = response.json()

for voice in catalog["data"]:
    # Send a built-in's id as the voice string; send others as {"id": ...}.
    print(f"{voice['type']:<8} {voice['language']}  {voice['id']}")
print(f"request_id: {response.headers.get('x-request-id')}")

JavaScript with fetch

Node.js 18 and later include fetch, which follows redirects by default.

const url = new URL("https://api.kittenml.com/v1/voices");
url.searchParams.set("model", "kitten-tts-2-latest");

const response = await fetch(url, {
  headers: {Authorization: `Bearer ${process.env.KITTENML_API_KEY}`},
});
const requestId = response.headers.get("x-request-id");
const catalog = await response.json();
if (!response.ok) {
  throw new Error(`${catalog.error?.message} (request_id=${requestId})`);
}

for (const voice of catalog.data) {
  // Send a built-in's id as the voice string; send others as {id}.
  console.log(`${voice.type.padEnd(8)} ${voice.language}  ${voice.id}`);
}
console.log(`request_id: ${requestId}`);

Response

This response is shortened to one built-in voice and one saved voice. A real response lists all 47 built-ins first, in number order, then the caller's saved voices, newest first, then any shared voices.

{
  "object": "list",
  "model": "kitten-tts-2-latest",
  "data": [
    {
      "id": "frank_gravelly_male_18",
      "object": "audio.voice",
      "name": "18 — Frank — Gravelly Male",
      "type": "built_in",
      "model": "kitten-tts-2-latest",
      "language": "en",
      "experimental": false
    },
    {
      "id": "voice_0123456789abcdef0123456789abcdef",
      "created_at": 1787760000,
      "name": "My saved voice",
      "object": "audio.voice",
      "type": "custom",
      "model": "kitten-tts-2-latest",
      "language": "en",
      "experimental": false
    }
  ],
  "public_voices_unavailable": false,
  "languages": [
    {"code": "en", "experimental": false},
    {"code": "ar", "experimental": false},
    {"code": "de", "experimental": false},
    {"code": "es", "experimental": false},
    {"code": "fr", "experimental": false},
    {"code": "it", "experimental": false},
    {"code": "pt", "experimental": false},
    {"code": "ru", "experimental": false},
    {"code": "zh", "experimental": false},
    {"code": "hi", "experimental": false}
  ],
  "voice_cloning": {
    "supported": true,
    "custom_voices_included": true,
    "saved_voices_supported": true,
    "supported_languages": ["en"],
    "create_endpoint": "/v1/audio/voices",
    "consent_endpoint": "/v1/audio/voice_consents"
  },
  "has_more": false,
  "first_custom_id": "voice_0123456789abcdef0123456789abcdef",
  "last_custom_id": "voice_0123456789abcdef0123456789abcdef"
}

Each entry in data is an audio.voice object:

Field	Meaning
`id`	The value to select this voice with
`name`	For a built-in, its title, such as `18 — Frank — Gravelly Male`; for a saved or shared voice, the name it was created with
`type`	`built_in` for the catalog below, `custom` for a saved voice your organization owns, or `public` for a shared voice
`model`	The model ID you sent
`language`	ISO 639-1 code of the language the voice speaks
`experimental`	`false` for every current voice
`created_at`	Unix timestamp in seconds; saved and shared voices only

Select a built-in by sending its id as the voice string. Select a saved voice with an object, "voice": {"id": "voice_..."}; a bare voice_... string is checked against the built-in catalog and rejected with 400 invalid_voice. On the input-streaming WebSocket, put either id in the voice query parameter.

Entries with type: "public" are shared voices that KittenML makes available to every organization. Select one by its name as the voice string, for example "voice": "Warm Narrator", or by its id exactly like a saved voice; on the WebSocket, either goes in the voice query parameter. Shared voice names are unique and never reuse a built-in voice ID, which always selects the built-in. They follow your own saved voices on every page and are not counted by limit or the cursors. public_voices_unavailable is true when the shared voices could not be read for this response; the built-ins and your saved voices are still complete.

Saved and shared voices currently use English reference transcripts and report language: "en".

Built-in voices

Kitten TTS 2 has 47 built-in voices. Select one by its voice ID: lowercase words and the voice's two-digit number, separated by underscores. Copy the ID from this table or from the response's id; the title is the voice's display name (name). Voice IDs are case-sensitive. The default voice is eleanor_somber_female_32.

Voice ID	Title	Language
`matthew_exhausted_mechanic_male_01`	01 — Matthew — Exhausted Mechanic Male	English
`willow_hushed_young_female_02`	02 — Willow — Hushed Young Female	English
`dolores_southern_elderly_female_03`	03 — Dolores — Southern Elderly Female	English
`victor_controlled_fury_older_male_04`	04 — Victor — Controlled Fury Older Male	English
`dante_playful_smooth_young_male_05`	05 — Dante — Playful Smooth Young Male	English
`alfred_tender_older_male_06`	06 — Alfred — Tender Older Male	English
`saoirse_joyful_irish_female_07`	07 — Saoirse — Joyful Irish Female	English
`claire_playful_professional_female_08`	08 — Claire — Playful Professional Female	English
`raven_controlled_fury_young_female_09`	09 — Raven — Controlled Fury Young Female	English
`bella_10`	10 — Bella (Kitten V0.8)	English
`jasper_11`	11 — Jasper (Kitten V0.8)	English
`luna_12`	12 — Luna (Kitten V0.8)	English
`bruno_13`	13 — Bruno (Kitten V0.8)	English
`rosie_14`	14 — Rosie (Kitten V0.8)	English
`hugo_15`	15 — Hugo (Kitten V0.8)	English
`kiki_16`	16 — Kiki (Kitten V0.8)	English
`leo_17`	17 — Leo (Kitten V0.8)	English
`frank_gravelly_male_18`	18 — Frank — Gravelly Male	English
`herbert_wise_male_19`	19 — Herbert — Wise Male	English
`diana_stern_female_20`	20 — Diana — Stern Female	English
`laurence_dramatic_male_21`	21 — Laurence — Dramatic Male	English
`maeve_cozy_female_22`	22 — Maeve — Cozy Female	English
`walter_warm_male_23`	23 — Walter — Warm Male	English
`edith_soft_female_24`	24 — Edith — Soft Female	English
`miles_gentle_male_25`	25 — Miles — Gentle Male	English
`grace_soothing_female_26`	26 — Grace — Soothing Female	English
`reginald_distinguished_male_27`	27 — Reginald — Distinguished Male	English
`iris_hushed_female_28`	28 — Iris — Hushed Female	English
`marcus_smooth_male_29`	29 — Marcus — Smooth Male	English
`serena_gentle_female_30`	30 — Serena — Gentle Female	English
`julian_dreamy_male_31`	31 — Julian — Dreamy Male	English
`eleanor_somber_female_32`	32 — Eleanor — Somber Female	English
`otis_rumbling_male_33`	33 — Otis — Rumbling Male	English
`vincent_hushed_male_34`	34 — Vincent — Hushed Male	English
`martha_weathered_female_35`	35 — Martha — Weathered Female	English
`sable_whispering_female_36`	36 — Sable — Whispering Female	English
`victoria_regal_female_37`	37 — Victoria — Regal Female	English
`arabic_38`	38 — Arabic (العربية)	Arabic
`german_39`	39 — German (Deutsch)	German
`english_40`	40 — English	English
`spanish_41`	41 — Spanish (Español)	Spanish
`french_42`	42 — French (Français)	French
`italian_43`	43 — Italian (Italiano)	Italian
`portuguese_44`	44 — Portuguese (Português)	Portuguese
`russian_45`	45 — Russian (Русский)	Russian
`chinese_simplified_46`	46 — Chinese (Simplified) (简体中文)	Simplified Chinese
`hindi_47`	47 — Hindi (हिन्दी)	Hindi

Voices 01–37 speak English. Voices 38–47 each speak the language in their name; select the language by selecting the voice. See Language coverage.

Voice IDs use only letters, digits, and underscores, so they need no encoding in a URL, such as the input-streaming WebSocket query.

KittenTTS 0.8 voices

The three KittenTTS 0.8 models (kitten-tts-nano-0.8, kitten-tts-micro-0.8, and kitten-tts-mini-0.8) share a fixed catalog of eight English voices. This endpoint does not list them; send the value from this table as voice. Voice values are case-sensitive, and the default is Bella.

Voice	Character
`Bella`	Warm and expressive
`Jasper`	Clear and conversational
`Luna`	Calm and smooth
`Bruno`	Deep and steady
`Rosie`	Bright and friendly
`Hugo`	Authoritative
`Kiki`	Lively and energetic
`Leo`	Relaxed and natural

A KittenTTS 0.8 voice is not accepted by Kitten TTS 2, and a Kitten TTS 2 voice is not accepted by a 0.8 model.

Saved-voice pagination

All 47 built-ins are returned on every successful page. limit controls only the number of caller-owned saved voices and accepts 1–100. When has_more is true, send last_custom_id as after:

curl -L --fail-with-body -sS \
  'https://api.kittenml.com/v1/voices?model=kitten-tts-2-latest&limit=25&after=voice_0123456789abcdef0123456789abcdef' \
  -H "Authorization: Bearer $KITTENML_API_KEY"

after must name a saved voice the caller owns. An unknown, revoked, or foreign ID is rejected with 404 voice_not_found and param: "after" rather than being treated as an empty page, so read the cursor from a previous response's last_custom_id instead of constructing it.

The endpoint requires a durable organization or registered legacy API key because its response can contain private saved-voice IDs. Another organization cannot list your voices. GET /v1/audio/voices is not a listing route and answers 405.

Create and revoke saved voices using the endpoints in Clone a voice. Runnable Python and JavaScript clients are available in the voice-listing TTS examples.

Query parameters​

cURL​

Python with httpx​

JavaScript with fetch​

Response​

Built-in voices​

KittenTTS 0.8 voices​

Saved-voice pagination​