How to Turn a Photo Into a VTuber Avatar

Most people who want a VTuber avatar don't draw. The usual answer is "commission an artist", which is the right call for a long-term character but a lot of money to spend on a maybe. An AI avatar from a photo is the cheap way to find out whether you like streaming as a character at all.

What happens when you turn a photo into an avatar?

The AI doesn't trace your face. It redraws you as a new cartoon character and keeps only general traits: hair style and colour, clothing, accessories. That's on purpose. Our AI Avatar Maker is told to keep "only general traits", so the result is an original character inspired by you, not a filter over your face.

To be useful for streaming, an avatar needs more than one picture. Our tool runs four steps:

That's about 37 seconds of model time, and it needs a plan. You get two transparent PNGs.

What photo works best?

We ran the same photo twice. Once as it was, well lit and 1024 pixels wide. Once darkened and shrunk to 384 pixels, like a grab from a dim webcam. The difference was bigger than we expected.

In the dark photo the model couldn't see the hair or the sweater, so it guessed. So before you upload:

• Face a window or a lamp. Light on your face beats a better camera.
• Frame your head and shoulders, looking at the camera.
• Wear the clothes and accessories you want the character to have. They carry over; your exact face doesn't.
• Use a photo at least 1000 pixels wide if you can.

Which style should you pick?

The tool has four styles: anime, 3D cartoon, pixel art and chibi. We also ran chibi on the good photo (15 seconds). It kept the hair and the green sweater but turned the long hair into a short bob, which chibi's big-head proportions tend to do. Anime kept the most detail.

How do you make the avatar talk?

1. 1

   Click Open in PNGTuber Maker

   When the frames are done, this button opens the PNGTuber Maker in the same browser with both pictures already in Idle and Talking. You can also save the PNGs and drop them in yourself; a blink frame is optional.

2. 2

   Press Start mic

   Allow the microphone when the browser asks.

3. 3

   Talk

   The mouth opens when your voice passes the threshold (12% by default). Raise it if background noise sets it off.

What are the limits?

• It won't look exactly like you. That's by design: you get an original cartoon character. If you want a realistic face moving live on camera instead, any face you choose, that's live face swap, a different tool.
• Two frames is a PNGTuber, not a rigged model. Your avatar talks and bounces but doesn't turn its head. For that you need a 3D model; see how to make a VTuber avatar.
• Text prompts that ask for celebrities or real public figures are blocked.
• Results vary run to run. If one comes out wrong, run it again with a better photo before changing styles.