ThumbCraftedTry Free Editor
Free · Private · No Upload

Fix Image Perspective

Shot a screen, a sign or a page at an angle? Put four corners on it and get it back flat.

Drag a handle onto each corner of the thing you photographed. The page solves the projective transform between those four points and a rectangle, then redraws every pixel through it, so straight lines stay straight and the far end is stretched back to its real size. It runs on your device.

Drop or click to upload

Drag & drop or paste from clipboard

JPG, PNG, WebP, 25MB max

No upload Touch-sized handles No watermark

Best for flat things shot at an angle: a laptop or TV screen, a whiteboard, a poster, a sign, a book page, a receipt, the front of a box. For a photo that is only tilted, rotate it instead.

1. Put a handle on each corner

Drag the four handles onto the corners. A magnifier follows your finger. Arrow keys nudge the focused handle.

2. The flattened result

How it works in three steps

1

Upload the photo

Drop, paste or pick a JPG, PNG or WebP up to 25 MB. It is read into your browser and stays there.

2

Drag the four corners

Put each handle on a corner of the screen, sign, page or box. The magnifier shows the exact pixel under your finger, and the result updates as you drag.

3

Pick a shape and save

Keep the shape measured from the corners or choose 16:9, A4 and others, then download a PNG or JPG or open it in the editor.

Why a photo of a screen looks wrong

A camera pointed at a flat object from an angle sees the near edge bigger than the far edge. The rectangle you photographed lands in the picture as a four-sided shape with no right angles: a trapezoid if you stood off to one side, something less regular if you were also above or below it. Rotating cannot fix that, because no single rotation makes all four edges parallel again. Cropping cannot fix it either, because the corners are in the wrong places relative to each other.

What fixes it is the transform a camera applied in the first place, run backwards. In maths it is called a homography: a 3 by 3 matrix that maps any four points on one flat plane to any four points on another, while keeping straight lines straight. Four corner pairs are exactly enough to pin down its eight unknowns. This page builds those eight equations from your four handles and the four corners of the output rectangle, solves them, and then for every pixel of the output works out which point of your photo it came from.

What happens to the pixels

The output is drawn backwards, one pixel at a time. For each pixel of the flat result, the matrix gives a position in the original photo, usually between pixels, and the colour is blended from the four photo pixels around that position (bilinear sampling). Drawing it this way round means every output pixel gets a value, with no gaps or holes, which is the problem with pushing source pixels forwards.

The far end of a skewed object had fewer pixels to begin with, so straightening it stretches those pixels. Text near the far edge will be softer than text near the camera. Nothing can restore detail the camera never recorded, but a straight result at the right proportions is far easier to read and far better in a thumbnail than the skewed original.

Getting the shape right

By default the output width is the longer of the top and bottom edges you marked, and the height is the longer of the two sides. That keeps every bit of resolution the photo has, but it is not always the true shape: the far edge of a skewed object is foreshortened, so a 16:9 screen shot steeply from the side can come out narrower than 16:9. If you know what the object is, pick its shape from the list. A TV or laptop screen is almost always 16:9, a sheet of paper is A4 or US Letter, a phone screen held upright is close to 9:16. The longer side stays at the size measured from your corners and the other side follows the shape you chose.

Using it for a thumbnail

A straightened screen is the cleanest way to show a game, an app or a website in a thumbnail without a screen recording. Shoot the monitor, flatten it here at 16:9, then open it in the editor to put your face and title over it. A flattened page or receipt works the same way for "I read the fine print" videos. If the result is soft, the sharpen tool brings back some edge contrast, and the Thumbnail Autopsy tells you whether the text still reads at feed size.

Perspective Questions

Is my photo uploaded?
No. The photo is decoded into a canvas in your browser, the transform is solved and applied there, and the download is built there. The line under the upload box counts the bytes of request body that leave the page while you work, and you can check the same thing in your browser's network panel.
Is this a real perspective correction or a stretch?
A real one. Stretching or skewing an image moves its corners but bends the lines in between. This page solves the 3 by 3 projective matrix that takes your four corners to a rectangle and maps every pixel through it, which is the transform a camera applies to a flat surface. Straight lines in the photo stay straight in the result.
Why is the far side of my result blurry?
Because the camera saw it smaller. The far edge of a skewed screen or page was recorded with fewer pixels, and straightening it spreads those pixels over more space. The fix is at capture time: stand more square to the object, or move closer. The sharpen tool can add back some edge contrast.
The result looks squashed. What happened?
The shape measured from your corners is only an estimate, because the far edge is foreshortened. Pick the object's real shape in the Shape list, such as 16:9 for a screen or A4 for a page, and the result is drawn at that shape instead.
Can it find the corners for me?
Not yet. The handles start just inside the edges of your photo and you place them. Snap to image edges puts all four on the photo's own corners, which is the right start when the object nearly fills the frame. Placing them yourself is also the most reliable way when the object has rounded corners, a bezel or a shadow.
Does it work on a phone?
Yes. Each handle has a touch area of 44 by 44 pixels, the photo area does not scroll while you drag, and a magnifier above your finger shows what is under it. On a desktop you can also focus a handle with Tab and nudge it with the arrow keys, one screen pixel at a time or ten with Shift.
What are the limits?
25 MB per image, in any image format your browser can open (JPG, PNG, WebP, GIF, BMP and others), one image at a time. The output is capped at 4,096 pixels on its longer side; a larger result is scaled down to that, and the size line under the preview says so. The four corners must form a convex shape. Download a PNG, or a JPG at 92% quality. There is no daily limit.
Does it keep my photo's metadata?
No. The result is a new image drawn on a canvas, so the camera model, date and GPS position in the original file are not copied into the download. If you only want to remove metadata from a photo without changing it, use the EXIF remover.
Is there a watermark?
No. Nothing is drawn on your image. The only mark is the file name, which starts with ThumbCrafted- and can be renamed.

Continue crafting

Keep going: your result carries over to the next tool with no re-upload.

  1. 1EditAdd text, effects, or a fresh background
  2. 2CropTrim or reshape to any aspect