Crop every face in a group photo, image or picture into a headshot of its own. Out comes one file per face, cut from the full-resolution photo at the ratio and head size you choose, with no face cut in half.
The face finder runs on this device after a one-time download whose size is stated on the button. Faces are found, not recognised: the boxes live in this tab, and each headshot can be adjusted by hand before it is saved.
One face, framed for a ratio: AI smart crop. A headshot for a form: passport photo crop or 2x2 photo crop. Avatars: PFP cropper. Many photos to one size: bulk crop.
What you can do here
One box per face
Each face gets a headshot box with the head at the share of the frame you choose (33–75%) and the eyes 38% down.
No one is sliced
A box that would cut a neighbour's face is shifted or tightened until that face is wholly out; the row says which it did.
Saved separately or zipped
Save one headshot, or all of them as a ZIP named photo-face-1, photo-face-2 and so on.
Found, not recognised
The face finder locates faces; it does not identify anyone, compare photos or keep anything after the tab closes.
How to face cropper
- 1Open the group photo
Drop it on the box. It stays in this tab.
- 2Find the faces
Press Find faces. The first time, the face finder downloads once (the size is on the button); after that it runs straight away.
- 3Choose the ratio and save
Pick 1:1, 4:5, 3:4 or 35:45 and the head size. Frame any one by hand with Frame it, save it, or save all as a ZIP.
Questions about cropping
How does the face cropper find faces?
It runs YuNet, a small face-detection model (MIT licence), on this device after a one-time download from our model file server, with the runtime from this site. It returns a box and five points per face; nothing is uploaded.
Does it recognise who is in the photo?
No. It finds where faces are, not whose they are. Nothing is compared across photos and no face data is kept: the boxes are numbers in this tab and are gone when you close it.
What does head size mean?
How much of the frame's height the head takes, from the chin to an estimated crown. 55% is a typical headshot; 75% is tight, 33% shows shoulders. The eyes are placed 38% down from the top.
What if two people are close together?
Each box is shifted sideways, and if that is not enough, tightened, until the neighbour's face is wholly outside it. If they are too close for either, the row is flagged so you can frame it by hand.
Which sizes do the files come out at?
At their own pixels: each headshot is cut from the full-resolution photo with no resize. The format and quality follow the Export settings; each row shows its size before and after saving.
A face was missed. Why?
Faces turned far away, very small, blurred or in deep shadow can be missed. Frame that one by hand: the crop box is still there and Download saves it.