PvcVoiceService
Upload PVC voice samples
Upload audio samples to a draft professional voice clone (PVC). Each request is all-or-nothing: if one file is rejected, none are saved.
/voices/v1/pvcVoices/{pvcVoiceId}/samplesYou can upload samples in several requests. They accumulate on the voice until you train it.
If an upload is rejected, for example because a file's format is unsupported or it is too large, the entire batch is dropped, including valid files. Split the batch and retry with only the accepted files.
The API detects the format from the file's bytes, not its declared MIME type or extension. Renaming a file to .wav does not make it a WAV file.
Accepted audio
| Format | Notes |
|---|---|
| WAV | PCM (8/16/24/32-bit), 32-bit float, or WAVE_FORMAT_EXTENSIBLE |
| MP3 | โ |
| WEBM | Must include a Duration element in its metadata |
Rejected: A-law WAV, 12-bit WAV, .m4a, zero-byte files, truncated WAV files, and WEBM files missing Duration metadata.
Limits
| Limit | Value |
|---|---|
| Files per request | 10 |
| Size per file | 1 GB |
| Combined sample storage per voice | 1 GB (1,000,000,000 bytes) |
| Samples per voice | 50 |
| Concurrent uploads per account | plan-dependent |
Exceeding a per-request, per-file, or per-voice limit returns HTTP error code 400. Too many concurrent uploads returns 429.
For recording and preparation tips, see Voice Cloning best practices.
Audio files to upload, up to 10 per request. Repeat the files[] form field once per file (e.g. files[]=@"sample-01.wav", files[]=@"sample-02.wav"). It is not a single field containing a list. Accepted formats: WAV, WEBM, MP3. The actual byte content is sniffed to determine format, not the declared MIME type or file extension. Each file is capped at 1,000,000,000 bytes (1 GB); the voice's combined sample storage is capped at 1,000,000,000 bytes (1 GB) and 50 samples total.
The samples accepted by this request, in the order they were uploaded.
Show child attributes
Size of the uploaded file, in bytes. Encoded as a string, like other 64-bit integers in this API.
Detected audio format, sniffed from the file's byte content.
Available options:audio/wavaudio/webmaudio/mpeg
Base64-encoded MD5 of the stored object, for verifying upload integrity against the source file.
Whether background-noise removal is applied to this sample during training, as set by remove_background_noise on upload.
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.
Available options:PVC_SAMPLE_ROLE_UNSPECIFIEDPVC_SAMPLE_ROLE_TRAININGPVC_SAMPLE_ROLE_GOLDENPVC_SAMPLE_ROLE_AVOID