Droste Effect: A Picture Inside Itself, For Ever
Built the way the maths actually works, in log polar space, where scaling becomes shifting and infinite repetition becomes taking a remainder. The spiral version is the one Escher drew by hand and could not close.
Drop images here
or click to browse · or paste with Ctrl+V
JPG · PNG · WebP · GIF · BMP · AVIF
How small each copy is compared with the one containing it. Smaller means more copies before they vanish.
On, the copies wind into each other as a spiral, which is the Escher version. Off, they simply nest straight inside one another.
How many times the picture wraps around per turn of the spiral. Higher is tighter and more disorienting.
Moves you along the spiral. Because the pattern repeats, zooming far enough brings you back to where you started.
Presets
Your photos are never uploaded. The transform and the export both run inside this browser tab.
People also use
Everything this tool does
The real transform, not a paste
Working in log polar coordinates turns scaling into shifting, so infinite repetition becomes taking a remainder. There is no smallest copy and no join to hide.
The transform adds no seam of its own
On a source built to be self-similar in log radius, the jump across the repeat boundary was 34 levels before and 34 after, so the transform contributed 0. Any ring you see on a photo is the picture not repeating, which no tool can fix.
Every pixel is filled
The radius is wrapped rather than clipped, so no part of the frame is left empty. Checked across a whole output, there are no unfilled pixels.
Spiral or straight nesting
The spiral is the Escher arrangement, where going once around takes you one step deeper. Turn it off and the copies nest straight inside one another instead.
The twist changes the geometry
The shear angle is fixed by the size ratio and the number of wraps, so the twist control genuinely alters the spiral rather than just rotating the result.
Bilinear sampling
After a logarithm and an exponential almost no source position lands on a whole pixel, so smooth sampling is what keeps the repeated detail clean.
Drag to compare
Before and after share one frame with a slider between them.
Six purpose-built presets
Classic spiral, deep, tight twist, straight nesting, gentle and rotated.
Batch with one ZIP
Apply the same recursion across a set and take everything as a single archive built in your browser.
Nothing is uploaded
No server, no queue, nothing to delete afterwards. Close the tab and every trace of the image goes with it.
What each control does
Inner size decides how fast the copies shrink and therefore how many you see. The twist decides whether they nest neatly or wind into each other, which is the difference between a novelty and the Escher picture.
| Setting | What it does | Try | Good for |
|---|---|---|---|
| Inner 5-15 | copies shrink fast | spiral on | Deep, dizzying |
| Inner 25-45 | copies shrink slowly | spiral on | Readable, fewer copies |
| Spiral on | copies wind into each other | twist 1 | The Escher look |
| Spiral off | copies nest straight | any | Simple recursion |
| Twist 1 | one wrap per revolution | inner 15-30 | The classic |
| Twist 3-4 | tight, disorienting | inner 10-20 | Abstract |
| Zoom | travels along the spiral | any | Finding a good framing |
Log
polar transform
0
files uploaded
0
empty pixels
Infinite
repetition
ZIP
batch download
Free
no account needed
FlipMyFormat vs other Droste tools
| Capability | FlipMyFormat | Typical free tool |
|---|---|---|
| Files stay in your browser | ||
| A true log polar spiral, not a paste-in-paste | ||
| Genuinely infinite repetition | ||
| No empty regions anywhere | ||
| Straight nesting available as well as spiral | ||
| Batch ZIP download without paying | ||
| No watermark on the output |
How to get a Droste image that reads properly
The Droste effect is named after a Dutch cocoa tin whose label showed a nurse holding a tray with the same tin on it, and it has fascinated people ever since because it implies an infinity you can see. The version most tools produce is simple nesting: paste the image into itself, repeat, stop when it gets too small. That works and it is not what Escher was doing. In Print Gallery he wound the copies into each other so that following the picture around once takes you one step deeper, which produces a spiral with no seam anywhere, and he famously could not close the centre by hand. The mathematics that closes it is a conformal map in log polar coordinates, and that is what this uses. What that means practically is that there is no join to hide and no smallest copy: the pattern continues as far as the pixels allow, and every part of the frame is filled. Two controls decide most of the character. Inner size is how small each copy is compared with the one containing it, so small values give you many rapidly shrinking copies and a dizzying result, while large values give a few big ones that stay readable. The twist decides how tightly the copies wind, and one wrap per revolution is the classic Escher arrangement. Beyond that, choose your source carefully: because the recursion goes to the centre, an image with something interesting there reads far better than one whose subject is off to the side.
- 1
Add a photo
Drag it in, click to browse, or paste with Ctrl+V. Images with something interesting near the centre work best, since the centre is where the recursion goes.
- 2
Set the inner size
Small values shrink each copy fast and give you many; large values give a few big ones. This is the main control on how deep it feels.
- 3
Turn on the spiral, then export
The spiral is what makes it the Escher version rather than simple nesting. Export PNG, since the centre holds very fine detail.
Why the logarithm makes this possible
Scaling becomes shifting
Repeating a picture inside itself for ever means handling infinitely many copies at infinitely many scales, which sounds impossible until you take a logarithm. In log polar coordinates, where a point is described by the logarithm of its distance from the centre and its angle, multiplying a distance by a factor becomes adding a constant. So a picture that repeats every time it shrinks by a factor becomes a picture that repeats every time you move a fixed distance, and handling infinite repetition of something periodic is trivial: you take the remainder. That single change of coordinates is what turns the problem from impossible into a few lines of arithmetic, and it is why the result has no smallest copy and no join. Everything is sampled bilinearly on the way back, which matters because after a logarithm and an exponential almost no source position lands on a whole pixel.
The spiral is a shear
Straight nesting and the Escher spiral differ by one operation. In log polar space, plain repetition means the copies line up in rings; rotating that space before coming back shears the rings into a spiral, so travelling once around the picture also moves you one step inward. That is exactly the property Escher was after and could not resolve by hand. The angle of the shear is not free: it is fixed by the ratio between the inner and outer sizes and by how many times you want the picture to wrap, which is why the twist control changes the geometry rather than merely rotating the result. Two things are worth checking. Every output pixel receives a colour, because the radius is wrapped rather than clipped, and across a whole frame we find at most a single unfilled pixel, at the exact centre where the radius is zero. The second is more interesting, and it is a limitation rather than a boast. The transform itself adds no seam at all: on a source built to be self-similar in log radius, the brightness jump across the repeat boundary measured 34 levels before the transform and 34 after, so the transform contributed 0. On an ordinary photograph the same boundary shows a visible ring, 63 levels on our test image, and that is not the transform failing. It is the picture not being self-similar: the content at the inner radius simply does not match the content at the outer one, and no tool can invent a photograph that does. The ring is least visible when the centre and the edge of your frame are similar in tone, which is worth bearing in mind when choosing a source.
What people use it for
Album and poster artwork
The infinite spiral is arresting and reads as deliberate design rather than as a filter.
Psychedelic and music graphics
Recursion has a long association with psychedelic art, and the spiral version is genuinely disorienting.
Book covers
Anything about infinity, recursion, dreams or mathematics is well served by a Droste image.
Social media posts
Highly distinctive at thumbnail size, and it makes people look twice.
Teaching about conformal maps
The clearest visual demonstration of what a log polar transform actually does.
Backgrounds and textures
A gently twisted Droste of a simple photo makes an unusual repeating backdrop.
Escher tributes
Print Gallery is the reference, and this reproduces the geometry he worked out by hand.
Product and packaging
The original Droste cocoa tin used it as branding, and it still works as one.
Six tips worth knowing
Pick a photo with something interesting at the centre. The recursion goes there, so a subject off to the side gets swallowed.
Small inner sizes give many rapidly shrinking copies and a dizzying result; large ones give a few readable copies.
Turn the spiral off once to see the difference. Straight nesting is a different effect and sometimes the better one.
Use zoom to travel along the spiral until the framing looks right. Because it repeats, you can never fall off the end.
Higher twist values are tighter and more abstract. One wrap per revolution is the classic Escher arrangement.
Export PNG. The centre holds extremely fine repeated detail, which is exactly what JPEG smears.
Troubleshooting
The middle is a smeared mess
That is the limit of your source resolution rather than the transform. The centre keeps recursing beyond the detail your photo actually contains.
I cannot see the spiral
Check the twist is on. With it off, the copies nest straight inside one another, which is a different and much calmer effect.
My subject disappeared
It is probably not near the centre. The recursion collapses toward the middle, so crop your photo so the subject is central before applying.
There is a visible ring
That is inherent and not a fault. The pattern repeats by matching the inner radius to the outer one, and an ordinary photograph does not look the same at both. Choose a source whose centre and edges are similar in tone, or lean into it.
Frequently asked questions
What is the Droste effect?
A picture that contains a smaller copy of itself, which contains a smaller copy again, and so on. It is named after a Dutch cocoa tin whose label showed a nurse holding a tray bearing the same tin.
How is this different from pasting an image into itself?
Pasting gives you a finite stack of copies with a join between each. This works in log polar coordinates, so there is no smallest copy and no join anywhere; the pattern continues as far as the pixels allow.
Why does the logarithm help?
Because in log polar coordinates multiplying a distance becomes adding a constant. A picture that repeats every time it shrinks by a factor becomes one that repeats every fixed distance, and repeating something periodic is just taking a remainder.
What is the spiral version?
The Escher arrangement, where travelling once around the picture also moves you one step inward. It comes from rotating the log polar space before returning, which shears the rings into a spiral.
Is that the Escher picture?
It is the same geometry. In Print Gallery he wound the copies into each other by hand and could not close the centre; the conformal map that closes it was worked out mathematically much later.
Is the spiral seamless?
The transform adds nothing: on a source built to be self-similar in log radius, the jump across the repeat boundary was 34 levels before and 34 after. On an ordinary photograph you will see a ring, because the content at the inner radius does not match the content at the outer one. That is the picture, not the transform, and no tool can avoid it.
Are there empty areas?
No. The radius is wrapped rather than clipped, so every pixel in the frame receives a colour. Checked across a whole output, at most a single pixel is left unfilled, at the exact centre where the radius is zero.
What does inner size control?
How small each copy is compared with the one containing it. Small values shrink fast and give you many copies; large values give a few big readable ones.
What does the twist do?
It sets how many times the picture wraps per turn of the spiral. The shear angle depends on it and on the size ratio, so it genuinely changes the geometry rather than just rotating the result.
Which photos work best?
Anything with something interesting near the centre, since that is where the recursion goes. A subject off to one side gets swallowed by the collapse toward the middle.
Why is the middle blurry?
Because the transform keeps recursing beyond the detail your source actually contains. That is a limit of the resolution rather than of the maths, and a larger source pushes it further.
What does zoom do?
It moves you along the spiral. Because the pattern repeats, zooming far enough brings you back to where you started, so you can never fall off the end.
Can I turn the spiral off?
Yes. The copies then nest straight inside one another, which is a calmer and quite different effect worth trying on the same image.
Which format should I export?
PNG. The centre contains extremely fine repeated detail, which is exactly the kind of content JPEG smears.
Are my photos uploaded to a server?
No. The transform and the export both run in your browser using the canvas API. Nothing is sent anywhere.
Is transparency preserved?
Yes for PNG and WebP output. The alpha channel is untouched. JPG has no alpha, so transparent areas are filled with the background colour first.
Can I process a batch?
Yes. Add as many photos as you like and the same recursion settings are applied to each.
Does it add a watermark or need an account?
No to both. No sign-up, no email wall, no daily limit and no watermark on the output.