> ## Documentation Index
> Fetch the complete documentation index at: https://docs.inworld.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Train a PVC voice

> Starts training on a draft voice's uploaded samples.
> 
> Returns the voice resource itself, **not** a long-running Operation. Poll the Get a PVC voice endpoint to track progress through `PVC_VOICE_STATE_QUEUED` → `PVC_VOICE_STATE_TRAINING` → `PVC_VOICE_STATE_READY` (or `PVC_VOICE_STATE_FAILED`).
> 
> Calling Train again while already queued or training is a no-op and returns the voice unchanged.

<Warning>
**This endpoint does not return an Operation.** There is no separate operation ID to look up. Poll [Get a PVC voice](https://docs.inworld.ai/api-reference/pvcAPI/pvcvoiceservice/get-pvc-voice.md) directly.
</Warning>

Training requires at least **600 seconds (10 minutes)** of sample audio in total, measured after any [trims](https://docs.inworld.ai/api-reference/pvcAPI/pvcvoiceservice/trim-pvc-voice-sample.md).

Your plan limits how often you can start training. Exceeding that rate returns `429`, with an error message that names the limit. Wait before trying again rather than retrying in a loop. See [Plan limits](https://docs.inworld.ai/tts/professional-voice-cloning.md#plan-limits) for each plan's allowance.

A voice in `PVC_VOICE_STATE_READY` cannot be trained again. The request returns `409`. To retrain, [create a new PVC voice](https://docs.inworld.ai/api-reference/pvcAPI/pvcvoiceservice/create-pvc-voice.md). See [Retrain or remove a trained voice](https://docs.inworld.ai/tts/professional-voice-cloning.md#retrain-or-remove-a-trained-voice).

## How long training takes

Training time depends on how much audio you uploaded, not how many samples it is split across. As a rough guide, an hour of audio takes about 10-30 minutes to train. It may take longer under heavy platform demand.

<Note>
Model training is capped at roughly 2800 seconds of audio and 1000 MB of source data per run. Audio beyond the cap is silently truncated rather than rejected. Keep total sample audio comfortably below the cap when uploading long recordings.
</Note>

## API reference

**Endpoint:** `POST https://api.inworld.ai/voices/v1/pvcVoices/{pvcVoiceId}:train`

### Authorization

- `Authorization` (string; required) — Your [API key](../../../api-reference/introduction). Read permissions are required for GET endpoints. Write permissions are required for POST, PATCH, and DELETE endpoints.
  
   For Basic authentication, please populate `Basic $INWORLD_API_KEY`. You can create a key in one command with the [Inworld CLI](../../../developer-tools/inworld-cli): `inworld workspace add-key`.

### Path parameters

- `pvcVoiceId` (string; required) — `pvcVoiceId` of the draft PVC voice to train.

### Request body

Content type: `application/json`


#### Request example

```json
{}
```

### Response

Status: `200`. Content type: `application/json`.

- `name` (string) — Resource name. Format: `workspaces/{workspace}/pvcVoices/{voice}`.
- `pvcVoiceId` (string) — PVC voice ID, derived from `displayName` at creation time. Use this value as `{pvcVoiceId}` on every other PVC endpoint. It is not the voice ID used for synthesis; see `customVoice`.
- `displayName` (string) — The human-readable name shown anywhere the voice is listed or selected.
- `description` (string) — Description of the voice.
- `tags` (string[]) — User-defined tags.
- `langCode` (enum<string>; options: "EN_US", "ZH_CN", "KO_KR", "JA_JP", "RU_RU", "AUTO", "IT_IT", "ES_ES", "PT_BR", "DE_DE", "FR_FR", "AR_SA", "PL_PL", "NL_NL", "HI_IN", "HE_IL") — Legacy enum form of the voice's language, such as `EN_US`. Returned alongside `languageCode`. Immutable after creation.
- `languageCode` (string) — The voice's language as a BCP-47-shaped locale string, e.g. `en-US`. Immutable after creation.
- `state` (enum<string>; options: "PVC_VOICE_STATE_UNSPECIFIED", "PVC_VOICE_STATE_DRAFT", "PVC_VOICE_STATE_QUEUED", "PVC_VOICE_STATE_TRAINING", "PVC_VOICE_STATE_READY", "PVC_VOICE_STATE_FAILED") — Lifecycle state of a PVC voice.
  
  - `PVC_VOICE_STATE_DRAFT`: Editable. Samples can be added, trimmed, or removed, and metadata can be updated.
  - `PVC_VOICE_STATE_QUEUED`: Training requested; waiting for a training slot.
  - `PVC_VOICE_STATE_TRAINING`: Actively training.
  - `PVC_VOICE_STATE_READY`: Training succeeded. Usable for TTS synthesis. It's permanent and cannot be deleted through this API.
  - `PVC_VOICE_STATE_FAILED`: Training failed. Editing the voice (e.g. renaming it, or adding/removing a sample) returns it to `PVC_VOICE_STATE_DRAFT` with its remaining samples intact.
- `failure` (object) — Populated on a PVC voice when its `state` is `PVC_VOICE_STATE_FAILED`.
  - `reason` (string) — Machine-readable failure code, such as `PVC_VOICE_FAILURE_REASON_INVALID_SAMPLES`. New values may be added over time, so don't validate against a hardcoded list. Fall back to displaying `message` for codes you don't recognize.
  - `message` (string) — Human-readable, scrubbed failure message. Never contains uploaded audio, filenames, or transcripts.
- `customVoice` (string) — Resource name of the trained voice, in the format `workspaces/{workspace}/voices/{voice}`. Set once the voice is `PVC_VOICE_STATE_READY`. Manage and use the trained voice through the Voices API.
- `samples` (object[]) — Audio samples currently attached to the voice.
  - `sampleId` (string) — Sample ID. Use this value as `{sampleId}` when trimming or deleting the sample.
  - `name` (string) — Resource name. Format: `workspaces/{workspace}/pvcVoices/{voice}/samples/{sample}`.
  - `fileName` (string) — Original file name of the uploaded audio.
  - `sizeBytes` (string) — Size of the uploaded file, in bytes. Encoded as a string, like other 64-bit integers in this API.
  - `durationSecs` (number) — Analyzed duration of the sample, in seconds, before any trim is applied.
  - `mimeType` (enum<string>; options: "audio/wav", "audio/webm", "audio/mpeg") — Detected audio format, sniffed from the file's byte content.
  - `hash` (string) — Base64-encoded MD5 of the stored object, for verifying upload integrity against the source file.
  - `removeBackgroundNoise` (boolean) — Whether background-noise removal is applied to this sample during training, as set by `remove_background_noise` on upload.
  - `trimStartMs` (integer) — Trim start offset in milliseconds, if set.
  - `trimEndMs` (integer) — Trim end offset in milliseconds, if set.
  - `role` (enum<string>; options: "PVC_SAMPLE_ROLE_UNSPECIFIED", "PVC_SAMPLE_ROLE_TRAINING", "PVC_SAMPLE_ROLE_GOLDEN", "PVC_SAMPLE_ROLE_AVOID") — How the sample is used in training. `PVC_SAMPLE_ROLE_UNSPECIFIED` is treated as `PVC_SAMPLE_ROLE_TRAINING` (ordinary training material). `PVC_SAMPLE_ROLE_GOLDEN` marks a positive reference whose character training should prefer; `PVC_SAMPLE_ROLE_AVOID` marks an advisory negative reference. New values may be added over time.
- `createTime` (string)
- `updateTime` (string)

### Response examples

#### 200: queued

```json
{
  "name": "workspaces/your_workspace_id/pvcVoices/my-professional-voice",
  "pvcVoiceId": "my-professional-voice",
  "displayName": "my-professional-voice",
  "langCode": "EN_US",
  "languageCode": "en-US",
  "state": "PVC_VOICE_STATE_QUEUED",
  "createTime": "2026-08-31T12:00:00Z",
  "updateTime": "2026-08-31T12:10:00Z"
}
```

#### 400: insufficient_audio

```json
{
  "code": 3,
  "message": "PVC samples must contain at least 600 seconds of audio",
  "details": []
}
```

#### 429: default_response

```json
{
  "code": 0,
  "message": "string",
  "details": [
    {
      "@type": "string"
    }
  ]
}
```

#### default: default_response

```json
{
  "code": 0,
  "message": "string",
  "details": [
    {
      "@type": "string"
    }
  ]
}
```

### Code examples

#### cURL

```bash
curl --location --request POST 'https://api.inworld.ai/voices/v1/pvcVoices/<pvc-voice-id>:train' \
--header "Authorization: Basic $INWORLD_API_KEY" \
--header 'Content-Type: application/json' \
--data '{}'
```

#### Python

```python
import requests

pvc_voice_id = "<pvc-voice-id>"
url = f"https://api.inworld.ai/voices/v1/pvcVoices/{pvc_voice_id}:train"
headers = {
    "Authorization": "Basic <api-key>",
    "Content-Type": "application/json"
}

response = requests.post(url, headers=headers, json={})
print(response.json())
```

#### JavaScript

```javascript
const pvcVoiceId = '<pvc-voice-id>';
const url = `https://api.inworld.ai/voices/v1/pvcVoices/${pvcVoiceId}:train`;

const response = await fetch(url, {
  method: 'POST',
  headers: {
    'Authorization': 'Basic <api-key>',
    'Content-Type': 'application/json',
  },
  body: JSON.stringify({}),
});

const data = await response.json();
console.log(data);
```