ThreeCutBlog
Try ThreeCut

How to capture a clean avatar

An avatar is a person you record once, then cast in any video by typing their @name. It needs about a minute of their voice and one clear photo. Ten minutes of care here is the difference between “that’s me” and “that’s a robot”.

The voice

You read six short lines, one at a time. The first is a consent line in your own voice. Each take is checked the moment you stop, for level, room noise and distortion, so you redo a bad one straight away.

a hand (15 cm)mica bitto the sidesoft, small: yeshard, echoey: no
  1. Pick a small, soft room. A bedroom or a closet full of clothes beats a kitchen or an office. Hard walls echo, and the echo gets cloned too.
  2. Make it quiet. Fans, air conditioning, the TV and notifications off; window closed.
  3. Mic a hand away, about 15 cm, a little to the side of your mouth so breaths don’t pop. Use the same mic for every line.
  4. Talk like you do on camera, not a reading voice. Natural energy, natural pace.
  5. Wait a beat before you start and after you finish. We trim the silence.
  6. About a minute is plenty. More than three minutes doesn’t make the voice better. No music, no other voices, no phone calls (they squash the sound).

The face

One photo is enough: you looking into the lens, head and shoulders, in soft light. Take it in the studio with your camera, or upload one.

eye levelan arm awaywindowhead and shoulderslook into the lensmouth closed, relaxednobody else in frame
  1. Face a window. Soft daylight on the face, never behind it. No hard shadows across the eyes.
  2. Camera at eye level, about an arm away. Head and shoulders in the frame.
  3. Look into the lens, face relaxed, mouth closed. The avatar moves the mouth itself.
  4. One person, plain background. No sunglasses or hats; everyday glasses are fine if they don’t glare.
  5. One clear photo is enough. Use the original, not a screenshot or a photo sent over a chat app (they lose detail).
  6. Only have group photos? Upload one and tap the person. Pick one where they face the camera and nobody overlaps them.

Someone else

You can make an avatar of a co-host, a teammate or a customer, with their agreement. They read the consent line themselves, in their own voice; a voice can’t be built without it. Avatars are private to your account, and deleting one deletes its voice and every file.

Using it

Type @name in a video’s topic: “@maya explains why sourdough fails”. The writer gives them the lines. With a photo, they appear on camera and speak; without one, they narrate over the pictures. In any scene you can also choose who says it.

How to make a clean AI avatar of yourself: voice and photo guide