Torn paper edge or paper fold — which one should you use?
A torn edge and a paper fold are two different mechanics, not two strengths of the same effect. A torn edge changes the outline of the piece: the shape is cut, so it reads as a scrap ripped out of a sheet. A fold keeps the rectangle and bends it in depth, so the outline stays a page and the crease does the work. This tool builds the first and not the second.
- Torn edge = the silhouette is cut. Fold = the silhouette stays and the surface bends. Different jobs, not two settings.
- The Cut style control here has three options — Torn, Deckle, Shredded. There is no fold style.
- The tear is synthesised from your image at fixed pixel wavelengths, not stamped from a template stretched to fit, so no two pieces share an outline.
- Choose the torn edge when the image should look ripped from something physical; choose a fold when a flat page should open.
A torn edge and a fold change different things
A torn edge is a silhouette operation. What you see is the image cut to a ragged outline, so the change happens at the boundary: the rectangle is gone, and what is left is a scrap with a raw paper core running along its rim. Everything that sells the effect — the uneven line, the exposed fibre, the shadow the scrap drops — hangs off that one cut.
A fold is a surface operation. The rectangle is still there; it is bent, and the bend reads because of shading and perspective. The outline moves a little in projection as the sheet turns, but the thing doing the work is the crease and the way light falls across it.
That difference decides which one you want, and it is worth stating plainly because the two are often sold side by side as if they were interchangeable. If the message is "this was ripped out of something physical", you want the torn edge. If the message is "a flat page is opening up", you want the fold. One is not a stronger version of the other.
It also decides what the tool has to be able to do. A silhouette can be drawn in two dimensions — no lighting model, no normals, just an outline and a fill. A fold needs a surface whose normal changes across it, which is a different piece of machinery entirely.
What actually makes an edge read as torn
The torn outline here is not a picture of a tear that gets scaled to fit your image. It is generated from the piece's own outline every time you load a photo.
The tool walks the rounded rectangle of the piece at a fixed step along its perimeter — about 4.6 px apart at the default Detail of 0.50, tightening toward 1.6 px as Detail rises — and pushes each sample along its outward normal.
The size of that push comes from value noise sampled by pixel position rather than by distance along the outline. That distinction is the whole trick. Keyed to pixels, the wavelengths stay put no matter how large the piece is; keyed to the outline, a long perimeter would stretch the same noise into a couple of slow lobes and you would get a bowed rectangle instead of a tear. Three octaves are summed: roughly 500 px, 180 px and 70 px, with amplitudes of 0.50, 0.18 and 0.06.
On top of that slow wander sits a per-sample jitter of a few pixels — about 1.2 + Detail × 4.5 px. It is added to the outer edge only, which means neighbouring samples are allowed to disagree. That disagreement is what reads as torn rather than as a smooth wobble; remove it and the edge turns into a gently wavy border.
Two numbers follow from the Edge thickness and Tear intensity sliders. The tear's reach is (3 + edgeWidth × 2.2) × (0.50 + intensity × 0.80) px — about 23.5 px at the defaults of 10 and 0.55. The raw paper core it exposes is (2 + edgeWidth × 0.90) × (0.50 + intensity × 0.60) px — about 9.1 px at the same settings. Both are pixels on the working canvas, so the look does not change with the image's original size.
The three edge styles, and what each does to the silhouette
The Cut style control offers three edges. The difference between them is mechanical rather than cosmetic — each one changes how the outline is built, not just how it is coloured.
- Torn (the default) runs the full noise stack at full gain. The outline wanders at all three wavelengths and carries the per-sample jitter, so it is the most irregular of the three.
- Deckle drops the shorter breaks and keeps only the long swell, and runs it at half gain. The result is the gentler edge a paper mill leaves on a sheet: fewer sharp teeth, a softer line, a shape that still reads as a cut edge rather than a tear.
- Shredded runs at 0.9 gain and adds a second, coarser tooth on top of the first. The silhouette comes out visibly rougher than Torn at the same Edge thickness and Tear intensity — closer to a torn-into-strips profile.
All three change the outer silhouette only. The inner edge — the line where the printed image meets the paper core — is built from the smooth wander alone, so the picture itself is never chewed up by the jitter. And none of the three is a fold: there is no style here that bends the piece in depth, because a silhouette renderer has no surface to bend.
When a fold is the better choice instead
Reach for a fold when the point is the page rather than the picture: a poster unrolling, a card opening, a newspaper page turning. The motion carries it, and a torn edge layered on top would only fight the crease.
Reach for a torn edge when the point is that the image came from somewhere: a clipping, a ransom-note scrap, a photo ripped off a wall.
If the fold is what you actually need, this tool is the wrong one. It renders a flat silhouette and has nothing to bend, so no setting here will get you there. Use a tool that models the sheet in three dimensions, or an editor with a fold preset, and bring the torn cutout in afterwards if you want both.
Choosing between them, in one line
Torn edge: the shape is the message — clippings, scraps, collage. Fold: the bend is the message — pages, cards, reveals. If you need both, they stack cleanly in an editor, because a torn cutout is just a piece you can lay on a page that something else is turning.
Questions people actually ask
Does this tool do a paper fold?
No. The Cut style control has three options — Torn, Deckle and Shredded — and none of them bends the piece. Everything on this page is a silhouette operation on a flat piece, so a fold is out of scope here.
What is the difference between Torn and Deckle?
Deckle keeps only the longest wavelength of the outline noise and runs it at half gain, so the edge swells gently instead of breaking into teeth. Torn uses all three wavelengths plus the per-sample jitter, which is what makes it read as ripped.
What does Shredded add on top of Torn?
A second, coarser tooth added per sample, at 0.9 gain. The silhouette comes out rougher than Torn at the same Edge thickness and Tear intensity, without touching the inner edge where the photo meets the paper core.
Why does the tear look the same on every image?
It should not. The outline is generated from a seed chosen fresh for each image, and Shuffle tear pattern re-seeds it on demand, so two pieces never share a tear. If two exports do look alike, the settings and the seed are the reason, not a fixed template.
Is the tear a template that gets stretched to fit my image?
No. It is synthesised from the piece's own outline by noise sampled at fixed pixel wavelengths, so the tear keeps its scale instead of being scaled with the image. A large photo and a small one get the same size of tear, not the same tear at different sizes.
Which slider changes the tear the most?
Edge thickness drives both how far the tear reaches and how wide the exposed paper core is; Tear intensity scales both at once. Detail changes the sampling step and the jitter, so it moves the texture of the edge more than its depth.
Try it on your own photo
Drop an image in, pick Torn, Deckle or Shredded, and export a still PNG or a WebM clip. Free, no account, and nothing is uploaded.
Open PaperRip