message.comDevelopers
Developers/Tutorials/Clone brand voice for AI

Clone your brand voice for AI

Upload a 15-second audio sample. We clone it into a custom voice that Smart Hold and AI Receptionist use on your phone lines. Free feature. The cloning runs through our neural voice synthesis pipeline; the voice profile lives in your workspace and is never shared across customers.

Voice cloning is part of the Phase C AI rollout. The API and UI flow described here will activate as the feature ships per workspace. See Voice cloning.

1. Record a 15-second sample

  1. Use a quiet room. Avoid keyboards, fans, traffic.
  2. Mic at a consistent distance. A USB condenser mic at 6 to 9 inches works.
  3. Speak naturally. Not slow, not fast. The voice you want the AI to sound like.
  4. Pick neutral content. Reading a paragraph of your About page is fine. A scripted welcome is fine.
  5. Export to .mp3 or .wav. 16-bit, 24 kHz or higher. Mono.

2. Upload

From the dashboard

  1. Open app.message.com.
  2. Go to Settings → AI → Voice profiles.
  3. Click Add voice profile.
  4. Name the voice (e.g., Acme Brand Voice), pick a personality (warm / professional / energetic / calm / custom), drop the audio file.
  5. Save.

From the API

bash
curl -X POST 'https://app.message.com/api/v1/voice-profiles' \
  -H 'Authorization: Bearer YOUR_WORKSPACE_JWT' \
  -F 'displayName=Acme Brand Voice' \
  -F 'personalityTag=warm' \
  -F 'sampleAudio=@/path/to/sample.mp3'

The response includes a providerVoiceId issued by the underlying voice synthesis provider. Cloning takes about 30 seconds.

3. Assign the voice to a phone number

  1. Open Settings → Phones → Numbers.
  2. Pick a DID.
  3. Under Voice profile, select the new voice.
  4. Save. The next call uses the cloned voice for Smart Hold and AI Receptionist.

4. Test

  1. Call the DID from your own phone.
  2. Trigger Smart Hold (let the call wait beyond the threshold) or AI Receptionist (press 1 at the IVR).
  3. Listen. The voice should match your sample.

Audio prep tips

  • Loudness. Target around -14 LUFS. Too quiet, the clone sounds thin; too hot, it clips.
  • Background noise. Anything continuous (HVAC hum) becomes part of the cloned voice. De-noise in Audacity or Adobe Audition first.
  • Music. Do not use audio with music. The cloner picks it up.
  • Diction. If you naturally have an accent, leave it. Cloning preserves it. Match the persona you actually want.

Common pitfalls

  • Public-figure voices. Cloning a famous person is prohibited by our acceptable-use policy. The system rejects obvious matches.
  • Consent. The voice must belong to someone who has consented. We require an attestation at upload time.
  • Voice drift. Cloned voices can drift on long generations. Smart Hold and AI Receptionist segment output so this rarely shows up, but be aware on multi-minute scripts.
  • Pronunciation of brand names. Cloned voices stumble on unusual brand names. Add the brand to Settings → AI → Pronunciation as a phonetic hint.

Next steps