Yes. Selfies, headshots and webcam stills all work. Run it through the four checks first: one face, even light, head and shoulders, and a JPEG, PNG or WebP up to 10 MB. What does not work is a screenshot of a video call, because it is soft before it starts.

My photo was turned down. What now?

Read the sentence you got. If it says we could not verify exactly one clear face, the photo has nobody in it, has more than one person, or the face is too dark or too small to read. Move to a window so the light is on your face, come closer so your head and shoulders fill the frame, and send it again. If it says the photo does not match the consented person, that twin was built from someone else's face, and a different likeness needs consent again.

Can I use a photo of somebody else?

Only if they build the twin, or give consent through the recording step themselves. A real person's face travels one way, as a Digital Twin, with consent on record. As a plain reference it is turned down, including your own face.

Do I need to crop or shrink it first?

No. Your browser compresses a phone photo before it is sent, so a big photo is not a problem. Cropping is worth doing only if the face is small in the frame, and even then the camera step in the Studio is faster than cropping by hand.

What happens to my photo after I send it?

It becomes the face your twin is built from, and it stays as you sent it. We do not use your photos, your voice, your scripts or your finished media to train a Kyndrify model. What we hold, and for how long, is written out in the biometric notice.

Can I change the photo later?

Yes. You can add more photos to a Digital Twin, and swap which one it is built from. If the twin already has a consent recording, a new photo is checked against that same face, so it has to be the same person. To use a different likeness, consent is given again.

Capture guide

How to take the photo everything is built from

Your Digital Twin is built from one photo, and every still, every look and every video comes from that one file. Here is how to take it, what the checks turn down, and what a render will accept when you hand it a reference later.

Your Digital Twin is built from one photo. Here is how to take it.

This is the content side. A photo of anyone but you needs their consent through the recording step, and your field's own advertising rules still apply to everything you record.

Part of Resources.

Checked September 2026. Free to read, no sign-up.

A woman in a grey hoodie films herself at her home desk beside a laptop.

What does a photo for an AI avatar need to be?

One person, facing the camera, in even light, with the face filling about the middle half of the frame. Send a JPEG, PNG or WebP up to 10 MB. We check for exactly one clear face and turn down the rest. It will not sharpen a blurry photo, and it will not accept a group shot.

Who it is for
Anyone about to build a Digital Twin, and anyone whose photo or reference file was just turned down.
What you need
One photo of yourself, taken on a phone, near a window. No studio, no lights, and no sign-up to read this.
What it costs
Nothing to upload. The video it feeds is 228 credits a second at 720p on every plan, and 540 a second at 1080p from the Plus plan up. The Free plan starts you with 8,400 credits, once.
What it does not do
It does not sharpen a blurry photo, it does not take a group shot, and a real person's face cannot ride through as a plain reference.

Most people search for AI avatar. We call yours a Digital Twin, because it is a real person with consent on record. An Avatar is a character nobody had to consent to.

One file, and everything after it

Most uploads are one input among many. This one is not. Your photo becomes the face in every still we make of you, every look you pick, and every video you render.

The recording you make next is checked against it, so if the photo is somebody else, the recording is turned down. Two minutes spent on it saves you doing the rest twice.

The framework

The four checks: face, light, frame, file

Run these four before you send anything. They are the four things that actually get a photo turned down.

Face
One person, looking at the camera, eyes visible. Not sunglasses, not a hat brim across the eyes, not a hand on the chin. Not two people, even if the second one is at the back of the shot. We check for exactly one clear face, and a photo with none, or with more than one, comes back.
Light
Even light on the whole face. Stand facing a window on a bright day, indoors, and you are done. What fails: a window behind you, midday sun overhead, a lamp to one side that leaves half the face dark, and a phone flash in a dim room.
Frame
Head and shoulders, with the face filling about the middle half of the frame. Close enough that the eyes read clearly, far enough that the top of the head is not cut off. Hold the phone at eye height, not below, and stand about an arm and a half away.
File
A JPEG, PNG or WebP up to 10 MB. You do not need to shrink it or crop it first: your browser compresses a phone photo on the way out, so the ceiling is almost never what stops you. Portrait or landscape both work.

A selfie is fine. A headshot is fine. A webcam still is fine. A screenshot of a video call is not, because it is soft before it starts.

Take it in the Studio, or bring one you already like

The Studio has a camera step with an outline on screen and one line of direction: look at the camera and give us a relaxed smile. Line your face up with the outline, hold still, and it takes the shot for you. That is the quickest route, and it gets the framing right without you thinking about it.

Bringing your own photo is just as good and sometimes better, because you already like it. Run it through the four checks first. If it passes, it goes through.

Either way, you can add more photos to a Digital Twin later, so the first one does not have to be the last word. The whole build, from the photo to the first video, is walked through on create an AI version of yourself.

What happens to your photo after you send it

Your photo stays exactly as you sent it. Nothing edits the original.

What we do is make new stills of you from it, in one of four levels. You pick the level before anything runs.

Keep my look
Your own hair and clothes, in a new room.
Light tidy
Freshly combed hair, simple clean clothes. This one is picked for you to start with.
Groomed
A fresh cut and a well-fitted shirt.
Styled looks
A new setting, with clothes to match.

Whichever you pick, your face stays exactly as it is. If the first set is not right, you can ask for another: you get three sets per twin while you are setting it up, and the cost of anything past that is shown before you press.

The recording is a second capture, and it has its own rules

After the photo you record a short clip. It does two jobs: it puts your consent on record, and it shows that the face in the photo is yours. The face in the recording is compared with the face in the photo, and a different person is turned down.

Length
No shorter than 15 seconds and no longer than 90 seconds. Shorter comes back as too short, longer as too long, and nothing past a minute and a half makes it better.
Sound
Speak clearly and a little louder than usual. Same light and framing as the photo, and somewhere quiet.
Retakes
Nothing is charged for recording, and you can record again any time.

Your Starter Twin’s short sample speaks in one of our Featured voices. Cloning your own voice is on every plan, Free included, and it needs ten seconds of clear audio: how voice cloning works.

References are a different job

Once your twin exists you will hand renders other files to work from: a picture of a room, a product shot, a clip whose motion you like. Those are references, and they follow a different set of rules from your photo.

How many
It depends on the style you pick. An image render takes up to four reference images, and a chosen character uses one of the four.
Files
JPEG, PNG or WebP up to 10 MB; MP4 or MOV up to 100 MB. Clips and audio are accepted on some styles and not on others.
Picture size
300 to 6,000 px on each side, between 2:5 and 5:2. A very long banner or a very tall strip falls outside it. If a file lands a little over, export it again at a smaller size and it goes through.
A clip does not carry a person
A clip you hand over as a reference guides the scene, the motion and the look. The person in it is not carried across, so if you upload a colleague expecting to see that colleague, you will not. That is what a talking-head video is for instead.

Read the per-style table

What each style accepts, how many, how large and how long.

What gets turned down, and why

Every reference goes through a quick content review first. It turns down these things:

Brand logos and trademarks
Turned down even when the mark is your own.
Copyrighted characters and artwork
A cartoon, a film still, a poster, a famous painting.
Real people's faces, unless they are a Digital Twin with consent
Including your own face.
Anything sexual, violent or hateful
Judged on the picture, not on what you meant by it.
Anyone under 18
Including in the background of a room shot.

Nothing is lost when it happens. Your credits come back, and the fix is usually small: swap the file, or crop out the part that caused it, and send it again.

The two that catch people out

Logos
The review sees the picture and nothing else, so it cannot tell your own mark from somebody else's. Two things work instead: describe the brand in words in your script, and use your own product photos shot with the mark out of frame.
Faces
A real person's face travels one way, as a Digital Twin, with that person's consent on record. A photo of a real person dropped in as a plain reference is turned down, and that includes your own face. Make the twin first, then render with it.

What a good photo does not do

It does not raise the resolution
Talking-head video renders up to 1080p, and 1080p starts on the Basic plan. A sharper photo does not buy a sharper video, so frame for the face rather than for the pixel count.
We do not retouch your original
Your file stays as you sent it. The look levels make new stills from it, and Keep my look keeps your own hair and clothes in a new room.
Your own face is not a plain reference
The Digital Twin path is the one way a real face travels, because it carries the consent with it. Dropping your own photo into a reference tray is turned down like anybody else's.
The re-rolls in the first run are a cap
Three sets of stills per twin while you set it up, and the door says no above that rather than quietly charging you.
A second twin is not free
The Free plan builds one Starter Twin, for life. More characters of your own start on the Plus plan, so adding another photo of a new person stops being free advice after the first twin. See what each plan includes.

Before you record somebody else

If the face is not yours, the person in the photo has to build the twin, or give consent through the recording step under their own name. Not a message saying it is fine. The record is the point.

When you publish, say that the video is AI generated where your platform or your field asks for it. From 2 August 2026 the EU AI Act asks that of the person publishing, and that is you, not us. Built-in AI labelling is still rolling out across output types, so write the line yourself rather than assuming the file carries it. The advertising rules of your own field still apply to every word you record, and following them stays yours. How we handle consent and disclosure.

This is not legal advice, and using Kyndrify does not make you compliant with anything.

Checked September 2026. This is a summary, not legal advice.

FAQ

Questions people ask with a camera in their hand

Can I use a selfie?: Yes. Selfies, headshots and webcam stills all work. Run it through the four checks first: one face, even light, head and shoulders, and a JPEG, PNG or WebP up to 10 MB. What does not work is a screenshot of a video call, because it is soft before it starts.
My photo was turned down. What now?: Read the sentence you got. If it says we could not verify exactly one clear face, the photo has nobody in it, has more than one person, or the face is too dark or too small to read. Move to a window so the light is on your face, come closer so your head and shoulders fill the frame, and send it again. If it says the photo does not match the consented person, that twin was built from someone else's face, and a different likeness needs consent again.
Can I use a photo of somebody else?: Only if they build the twin, or give consent through the recording step themselves. A real person's face travels one way, as a Digital Twin, with consent on record. As a plain reference it is turned down, including your own face.
Do I need to crop or shrink it first?: No. Your browser compresses a phone photo before it is sent, so a big photo is not a problem. Cropping is worth doing only if the face is small in the frame, and even then the camera step in the Studio is faster than cropping by hand.
What happens to my photo after I send it?: It becomes the face your twin is built from, and it stays as you sent it. We do not use your photos, your voice, your scripts or your finished media to train a Kyndrify model. What we hold, and for how long, is written out in the biometric notice.
Can I change the photo later?: Yes. You can add more photos to a Digital Twin, and swap which one it is built from. If the twin already has a consent recording, a new photo is checked against that same face, so it has to be the same person. To use a different likeness, consent is given again.

Keep going

You have the photo. Take it.

Four checks, two minutes, and the rest of it follows from that one file. The Free plan starts you with 8,400 credits and builds one Starter Twin. What we hold when you build one is written out in the biometric notice.

Read the guides Get started for free