Clone a voice from an audio sample and generate speech with full control over speed, pitch, emotion, and more.

Voice Cloning

Clone a custom voice from an audio sample, then use it to generate speech with full control over speed, pitch, emotion, and more.

Overview

Voice cloning is a two-step process:

Clone a voice — Upload an audio sample to create a custom voice ID.
Generate speech — Use the cloned voice ID to synthesize speech from text.

Step 1: Clone a Voice

Endpoint

POST /api/v1/voice-training

Audio Requirements

Requirement	Value
Minimum duration	10 seconds
Maximum file size	20 MB
Supported formats	`.mp3`, `.wav`
Recommended	Clear speech, minimal background noise

Request

Headers:

Header	Required	Description
`Authorization`	Yes	`Bearer YOUR_API_KEY`
`Content-Type`	Yes	`application/json`

Body:

{
  "audio_url": "https://example.com/my-voice-sample.mp3",
  "name": "My Custom Voice",
  "noise_reduction": true,
  "volume_normalization": true,
  "accuracy": 0.8,
  "preview_text": "Hello, this is a preview of my cloned voice."
}

Field	Required	Type	Description
`audio_url`	Yes	string	URL to the audio file (must be publicly accessible).
`name`	No	string	Name for the voice. Auto-generated if not provided.
`noise_reduction`	No	boolean	Enable noise reduction. Default: `false`.
`volume_normalization`	No	boolean	Normalize volume levels. Default: `false`.
`accuracy`	No	number	Text validation accuracy (0–1). Default: `0.8`.
`preview_text`	No	string	Custom text for the preview audio clip.

Response

{
  "success": true,
  "voice_id": "a1b2c3d4-e5f6-7890-abcd-ef1234567890",
  "name": "My Custom Voice",
  "preview_url": "https://storage.verbatik.com/audio/preview-abc123.mp3",
  "cost_cents": 300,
  "balance_cents": 1700
}

Cost: $3.00 per voice.

Step 2: Generate Speech with a Cloned Voice

Endpoint

POST /api/v1/voice-cloning

Headers

Header	Required	Description
`Authorization`	Yes	`Bearer YOUR_API_KEY`
`Content-Type`	Yes	`text/plain`
`X-Voice-ID`	Yes	The voice ID from the cloning step.
`X-Store-Audio`	No	`true` to store audio and get a URL. Default: `false`.

Voice Settings (optional headers)

Header	Range	Default	Description
`X-Speed`	0.5–2.0	1	Speech speed multiplier.
`X-Volume`	0–10	1	Volume level.
`X-Pitch`	-12 to 12	0	Pitch adjustment in semitones.
`X-Emotion`	See below	—	Emotion of the generated speech.
`X-English-Normalization`	true/false	—	Improve number and abbreviation reading.

Voice Modification (advanced, optional)

Header	Range	Description
`X-Voice-Modify-Pitch`	-100 to 100	Fine-grained pitch adjustment.
`X-Voice-Modify-Intensity`	-100 to 100	Energy/intensity of the voice.
`X-Voice-Modify-Timbre`	-100 to 100	Tonal quality adjustment.

Audio Settings (optional)

Header	Values	Default	Description
`X-Sample-Rate`	8000, 16000, 22050, 24000, 32000, 44100	32000	Sample rate in Hz.
`X-Bitrate`	32000, 64000, 128000, 256000	128000	Audio bitrate.
`X-Format`	mp3, pcm, flac	mp3	Output format.
`X-Language-Boost`	Language code	—	Enhance language recognition.

Body: Plain text to synthesize (max 5,000 characters).

Supported Emotions

happy, sad, angry, fearful, disgusted, surprised, neutral

Special Text Features

Interjection tags — Insert natural speech sounds: (laughs), (sighs), (coughs), (clears throat), (gasps), (sniffs), (groans), (yawns)

Pause markers — Insert precise pauses: <#x#> where x is the duration in seconds (0.01–99.99). Example: Hello <#1.5#> world

Pricing

$0.08 per 1,000 characters ($80 per 1M characters).

Example

curl -X POST https://api.verbatik.com/api/v1/voice-cloning \
  -H "Authorization: Bearer vbt_your_api_key" \
  -H "Content-Type: text/plain" \
  -H "X-Voice-ID: a1b2c3d4-e5f6-7890-abcd-ef1234567890" \
  -H "X-Speed: 1.1" \
  -H "X-Emotion: happy" \
  -H "X-Format: mp3" \
  --data "Hello! This is my cloned voice speaking with a happy emotion." \
  --output cloned-speech.mp3

Managing Cloned Voices

List Your Cloned Voices

GET /api/v1/my-voices

Parameter	Type	Description
`status`	string	Filter: `pending`, `ready`, or `failed`.

[
  {
    "id": "a1b2c3d4-e5f6-7890-abcd-ef1234567890",
    "name": "My Custom Voice",
    "status": "ready",
    "preview_url": "https://storage.verbatik.com/audio/preview-abc123.mp3",
    "source_audio_url": "https://example.com/my-voice-sample.mp3",
    "created_at": "2025-01-15T10:30:00.000Z",
    "last_used_at": "2025-01-20T14:00:00.000Z"
  }
]

Voice Statuses

Status	Description
`pending`	Voice is being processed (typically a few seconds).
`ready`	Voice is ready for TTS generation.
`failed`	Cloning failed. Check the error message.

Voice Expiration

Cloned voices expire after 7 days of inactivity. Verbatik automatically sends keep-alive requests for actively used voices. If a voice expires, you'll need to clone it again.

Uploading Audio

If your audio isn't hosted at a public URL, upload it first:

POST /api/audio-upload

Upload as multipart form data. The response includes a URL for the audio_url field.

Tips for Best Results

Use clear, high-quality audio — Record in a quiet environment.
Speak naturally — Natural speech patterns produce better clones.
Provide at least 10 seconds — Longer samples (30–60 seconds) generally produce better results.
Enable noise reduction — If your audio has background noise.
Use volume normalization — Helps with inconsistent audio levels.

Voice Cloning

On this page