Clone a voice

POST/v1/custom-voices

Clone a voice from a recording and use its voice ID to generate speech.

1. Upload your recording

Send multipart/form-data with file, a unique name, provider, and consent=true to confirm the speaker's permission and your rights to use the recording.

The examples use a WAV file named sample.wav. Replace it with a clean recording of one speaker, without music or background noise. Other accepted audio formats depend on the provider. See the sample requirements below before uploading.

2. Get the voice ID from the response

The response returns your voice ID in id:

{  "id": "AM:@narrator",  "voices": [{    "voice": { "id": "AM:@narrator", "status": "active" },    "model": { "id": "soniox/tts-rt-v2" },    "providers": [{ "status": "active" }]  }]}

In this example, the voice ID is AM:@narrator. Copy the complete value. You will use it to generate speech and to get or delete the voice later.

When both status values are active, the voice is ready. If either is still pending, wait and get the voice again. If cloning failed, failure_reason explains why.

3. Generate speech

Use id from the response as voice, and voices[].model.id as model:

{  "model": "soniox/tts-rt-v2",  "voice": "AM:@narrator",  "input": "Hello from my cloned voice."}

The complete code examples below upload your recording, check that your voice is ready, and save the generated audio to speech.mp3.

Sample requirements

Provider	Maximum sample size	Duration	Language field
`soniox`	10 MiB	At most 120 seconds	Omit
`fish`	10 MiB	Not specified	Omit
`minimax`	10 MiB	10–300 seconds	Omit
`cartesia`	10 MiB	At least 10 seconds	Required
`fal`	10 MiB	At least 10 seconds	Omit
`inworld`	10 MiB	3–30 seconds	Required

The complete multipart request must fit within 11 MiB. Use List providers to find providers with voice_cloning support. A cloning fee may apply; speech generation is billed separately.

Language codes

Use one of the codes below when sending language. Omit the field for providers marked Omit above. Where listed, auto asks the provider to detect the language.

cartesia: en, fr, de, es, pt, zh, ja, hi, it, ko, nl, pl, ru, sv, tr, tl, bg, ro, ar, cs, el, fi, hr, ms, sk, da, ta, uk, hu, no, vi, bn, th, he, ka, id, te, gu, kn, ml, mr, pa, or, ur
inworld: auto, ar, de, en, es, fr, he, hi, it, ja, ko, nl, pl, pt, ru, zh

Optional recording fields

Omit capture_script_version unless your application tracks versions of a recording script. Omit consent_id for the consent=true flow; it is only used with a separate spoken-consent challenge.

Authorization

BearerAuth

AuthorizationBearer <token>

Tenant key supplied as the Authorization Bearer token.

In: header

Request Body

multipart/form-data

TypeScript Definitions

Use the request body type in TypeScript.

file*file

Upload the audio file itself, not a URL or base64 string. See the provider-specific sample requirements above.

Formatbinary

name*string

Short name for the voice, unique within your organization. Lower case letters, digits and hyphens.

Length1 <= length <= 64

provider*string

Provider to clone the voice on. Use a provider whose voice_cloning feature is listed as supported on GET /v1/providers. See the sample requirements above.

Length1 <= length <= 64

language?string

Language code for the sample. Required for providers marked Required in the table above; omit for providers that infer it. Use auto only where listed.

Length1 <= length <= 35

description?string

Optional description to help you identify the voice.

Lengthlength <= 500

capture_script_version?string

Optional label for your recording script. Omit if you do not use a versioned script.

Match^[A-Za-z0-9][A-Za-z0-9_.-]{0,63}$

consent?string

Send the string true to confirm the speaker consented and you have the rights to clone the recording.

consent_id?string

Only for a provider requiring a separate spoken-consent challenge. Omit for the consent=true flow.

Lengthlength <= 120

Response Body

`application/json`

import { readFile, writeFile } from "node:fs/promises";// Node.js 22+. Set ALLMODELS_API_KEY and use your own sample.wav.const form = new FormData();form.set("file", new Blob([await readFile("sample.wav")], { type: "audio/wav" }), "sample.wav");form.set("name", "narrator");form.set("provider", "soniox");form.set("consent", "true");const response = await fetch("https://api.allmodels.io/v1/custom-voices", {  method: "POST",  headers: { Authorization: "Bearer " + process.env.ALLMODELS_API_KEY },  body: form});if (!response.ok) throw new Error("Clone failed (" + response.status + "): " + await response.text());const result = await response.json();const voiceId = result.id;console.log("Your voice ID:", voiceId);// If the voice is still pending, the speech call fails; get the voice again to check status and failure_reason.const speech = await fetch("https://api.allmodels.io/oai/audio/speech", {  method: "POST",  headers: {    Authorization: "Bearer " + process.env.ALLMODELS_API_KEY,    "Content-Type": "application/json"  },  body: JSON.stringify({    model: "soniox/tts-rt-v2", voice: voiceId,    input: "Hello from my cloned voice.", response_format: "mp3"  })});if (!speech.ok) throw new Error("Speech failed: " + await speech.text());await writeFile("speech.mp3", Buffer.from(await speech.arrayBuffer()));

import osimport requests# Set ALLMODELS_API_KEY and use your own sample.wav.with open("sample.wav", "rb") as audio:    response = requests.post(        "https://api.allmodels.io/v1/custom-voices",        headers={"Authorization": f"Bearer {os.environ['ALLMODELS_API_KEY']}"},        files={"file": ("sample.wav", audio, "audio/wav")},        data={"name": "narrator", "provider": "soniox", "consent": "true"},        timeout=120,    )response.raise_for_status()result = response.json()voice_id = result["id"]print("Your voice ID:", voice_id)# If the voice is still pending, the speech call fails; get the voice again to check status and failure_reason.speech = requests.post(    "https://api.allmodels.io/oai/audio/speech",    headers={"Authorization": f"Bearer {os.environ['ALLMODELS_API_KEY']}"},    json={"model": "soniox/tts-rt-v2", "voice": voice_id,          "input": "Hello from my cloned voice.", "response_format": "mp3"},    timeout=60,)speech.raise_for_status()with open("speech.mp3", "wb") as output:    output.write(speech.content)

# Requires jq. Set ALLMODELS_API_KEY and use your own sample.wav.set -ecurl --fail-with-body https://api.allmodels.io/v1/custom-voices \  -H "Authorization: Bearer $ALLMODELS_API_KEY" \  -F "file=@sample.wav;type=audio/wav" \  -F "name=narrator" \  -F "provider=soniox" \  -F "consent=true" \  --output cloned-voice.json# If the voice is still pending, the speech call fails; get the voice again to check status and failure_reason.jq '{model: "soniox/tts-rt-v2", voice: .id, input: "Hello from my cloned voice.", response_format: "mp3"}' \  cloned-voice.json > speech-request.jsoncurl --fail-with-body https://api.allmodels.io/oai/audio/speech \  -H "Authorization: Bearer $ALLMODELS_API_KEY" \  -H "Content-Type: application/json" \  --data @speech-request.json \  --output speech.mp3

{  "id": "AM:@narrator",  "voices": [    {      "model": {        "id": "soniox/tts-rt-v2",        "name": "Soniox tts-rt-v2"      },      "voice": {        "id": "AM:@narrator",        "name": "narrator",        "description": null,        "category": {          "id": "cloned",          "name": "Cloned"        },        "languages": [],        "scope": {          "type": "organization"        },        "status": "active",        "fee_credits": 0,        "first_seen_at": "2026-09-11T02:00:00.000Z",        "last_seen_at": "2026-09-11T02:00:05.000Z"      },      "providers": [        {          "id": "soniox",          "name": "Soniox",          "provider_model_id": "tts-rt-v2",          "status": "active",          "failure_reason": null        }      ]    }  ],  "total_count": 1,  "has_more": false,  "next_cursor": null}

{  "error": "invalid_request"}

{  "error": "invalid_api_key"}

{  "error": "insufficient_credits"}

{  "error": "tenant_disabled"}

{  "error": "name_taken",  "voice_id": "AM:@narrator"}

{  "error": "sample_too_large"}

{  "error": "consent_required"}

{  "error": "custom_voices_failed",  "voice_id": "AM:@narrator",  "message": "Creation could not be confirmed. Retrieve this voice using voice_id and check its status before uploading another sample."}

{  "error": "clone_failed"}

{  "error": "custom_voices_unavailable"}

1. Upload your recording

2. Get the voice ID from the response

3. Generate speech

Sample requirements

Language codes

Optional recording fields

Authorization

Request Body

Response Body

201application/json

400application/json

401application/json

402application/json

403application/json

409application/json

413application/json

422application/json

500application/json

502application/json

503application/json

`application/json`

`application/json`

`application/json`

`application/json`

`application/json`

`application/json`

`application/json`

`application/json`

`application/json`

`application/json`

`application/json`