How to clone yourself with AI: your face and your voice

To clone yourself with AI, you copy two things: your voice and your face. A voice clone is a synthetic voice built from a recording of you, and a lip sync model animates a photo of you to speech in that voice, so you appear to say lines you never recorded. On Sume you clone the voice in the app, speak the script with TTS 1.0, and lip sync your photo with VEED Fabric 1.0.

Facts come from Sume's Create new avatar and Models overview docs and the request schemas in the Sume API reference, read on 2026-09-27. The Voices steps and anything called current behavior are read from Sume's code. Clone only yourself, or someone who has agreed to it.

What do I need to clone myself?

• A photo of your face at a public HTTPS URL, for the lip sync step.
• A recording of your voice. In current code the clone upload takes a wav, mp3, m4a, mp4, webm, or ogg file of up to 20 MB, and the in-app recorder stops at 30 seconds.
• A Sume API key for the speech and lip sync jobs.

How do I clone my voice?

In the Sume app today, open Assets → Voices, choose Add a new voice, then Clone a voice, and upload or record your audio. Give the voice a name and pick its gender and its language. Copy ID gives its voi_ id, which TTS accepts as voice.id once the voice is ready; before that, current code answers 400 invalid_voice_id.

Cloning happens in the app, not over the API. AI voiceover in your own voice covers this step in more detail.

How do I make my clone say a new script?

Two jobs turn a script into a video of you saying it:

• Speak the script with POST /v1/tts-1.0/generate: the transcript, voice.id set to your voi_ id, and language if the script isn't English. With timestamps.words: true, current code reports the audio's duration_seconds beside its audio_url.
• Lip sync with POST /v1/veed/fabric-1.0: your photo as image_url, the TTS file as audio_url, and its length as duration_seconds, 1–300 seconds. The make a photo talk guide shows this request.

curl -X POST https://api.sume.com/v1/tts-1.0/generate \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: my-clone-line-001" \
  -d '{
    "transcript": "Quick update from me: the new plan starts on Monday.",
    "voice": { "id": "voi_…" },
    "language": "en",
    "timestamps": { "words": true }
  }'

Should I make an Avatar 1.0 avatar of myself instead?

An avatar gives you a reusable handle, but it is not a copy of you:

• From a photo, current code redraws you with an image model, prompted "Photo of this person", so the avatar is a generated likeness rather than your picture.
• Its voice is generated as well: in current code, a sample made to match the person's look, then cloned as English. An Avatar 1.0 talking video takes no voice field, so that video can't use your clone.
• In current code, the photo must be PNG, JPEG, WebP, or GIF, at least 64 px wide and tall, with the long side at most 6 times the short one; errors suggest a clear, front-facing image.
• Creation costs $0.95 per avatar. You can still pair the avatar with your voice: send only voice.id to TTS, then the avatar's avatar_handle to Fabric. A TTS request carrying both an avatar and a different voice.id fails with 400.

Can I make a digital twin from a video?

Not in one step: avatar creation takes a prompt, a profile, or a photo, and no video. Use the recording twice instead. Take a clear frame of your face as the photo, and clone the voice from the recording's audio, saved as a clip within the upload limits above. The result is animated from that one frame, not from the recording's own motion.

What are the limits, and what does it cost?

TTS costs $0.0475 per 1,000 characters and Fabric $0.1875 per audio second (720p), each plus a 5.5% agent fee by default. Fabric counts audio seconds, rounded up.