Custom Voice from Voice Design

Create a custom voice in two phases: generate candidate samples, then promote the one you like.

1. Generate Samples

curl 'https://app.resemble.ai/api/v2/voice-design' \
  -H 'Authorization: Bearer YOUR_API_TOKEN' \
  -F 'user_prompt=A middle-aged female with an Australian accent. She speaks with confidence and warmth at a moderate pace. Her tone is friendly and approachable, like a knowledgeable tour guide.'

Response:

{
  "voice_candidates": [
    { "audio_url": "...", "voice_sample_index": 0, "uuid": "abcd1234" },
    { "audio_url": "...", "voice_sample_index": 1, "uuid": "abcd1234" },
    { "audio_url": "...", "voice_sample_index": 2, "uuid": "abcd1234" }
  ]
}

Each candidate shares the same uuid (the design request) and exposes a unique voice_sample_index (0, 1, or 2). Listen to the audio URLs and pick your favorite.

2. Create the Voice

curl 'https://app.resemble.ai/api/v2/voice-design/abcd1234/1/create_rapid_voice' \
  -H 'Authorization: Bearer YOUR_API_TOKEN' \
  -F 'voice_name=Australian Female'

Response:

{ "voice_uuid": "1234567890" }

The returned voice_uuid is the actual voice asset; Resemble begins building it immediately.

3. Monitor Build Status

curl 'https://app.resemble.ai/api/v2/voices/1234567890' \
  -H 'Authorization: Bearer YOUR_API_TOKEN'

The status field transitions from initializing → running → finished. Use the Voices API to check component readiness or visit the dashboard.

4. Synthesize Audio

Once finished, use the voice with standard synthesis endpoints:

curl --request POST 'https://f.cluster.resemble.ai/synthesize' \
  -H 'Authorization: Bearer YOUR_API_TOKEN' \
  -H 'Content-Type: application/json' \
  --data '{
    "voice_uuid": "1234567890",
    "data": "Hello from Resemble!",
    "sample_rate": 48000,
    "output_format": "wav"
  }'

Use the resulting audio_content (base64) or audio_src URL to retrieve the audio.

Troubleshooting

• Prompts lacking detail may produce generic results—see How to Prompt.
• If builds remain running for longer than expected, contact support with the voice_uuid.