
Pick a photo with a clean outline
A subject that separates from its background — a pet against a plain floor, a face against a wall, an object on a table. Embroidery has a physical edge, and the model needs to be able to find one.
One photo · Satin stitch · A patch you could sew on
The stitched-patch look is everywhere on merch and nowhere in photo apps, because faking it with a texture overlay never survives a second glance. This redraws the subject in thread instead, one stitch direction at a time.
A run we did ourselves on this page, on a dog with an awkward coat. Drag the divider and check the markings and the collar tag first, then look at how the stitch rows change direction across the chest. Every output is AI-generated, so treat this as what the tool does rather than a result you will get back identically.


Drag to compare before and after
Most craft filters fail at the same place: they simulate the surface and ignore the construction.
Embroidery is not a texture, it is a set of decisions. Where does the fill start and stop. Which direction does each block of satin stitch run. Where does the thread change colour, and does it change abruptly the way real thread does or fade the way a gradient does. Which details are simply too small to stitch at this size and have to be dropped.
A filter that skips all of that and multiplies a canvas texture over your photo produces something that looks like a photo behind a mesh, which is what nearly every stitch effect on the app stores actually is.
The run above is ours, and the corgi is the useful test case because fur is the hardest thing to stitch convincingly. Look at where the stitch rows change angle across the chest and around the muzzle: they follow the way the coat lies, which is the decision a digitiser makes by hand before a machine ever runs.
Then look at the collar tag — small, dark, and still there, because it was worth the thread. Nothing about the pose, the markings or the ear position moved.
Two limits worth knowing. This produces a picture of an embroidered patch, not a machine file — there is no . dst or . pes to feed a Brother, and a digitiser would still have to do that work from your image. And fine detail simplifies on the way in: thin whiskers, small text and delicate jewellery tend to be absorbed into the fill, the same compression that happens in real stitching.
If you want the fabric itself changed rather than the whole subject converted, the clay filter is the neighbouring idea in a different material.
Creative engine
Upload a photo, pick the backing, and compare the stitch direction against the original.
One subject, clear outline, plain background. A busy scene has no edge to run a border around, and the border is half the effect.
Oatmeal linen, dark denim, a cap panel. The material behind the patch changes the whole mood and costs you nothing to specify.
The photo you choose does most of the work, and the choices are not the ones you would make for a portrait.

A subject that separates from its background — a pet against a plain floor, a face against a wall, an object on a table. Embroidery has a physical edge, and the model needs to be able to find one.
Linen, denim, a canvas tote, a dark cap. Naming the backing gives the light something to behave against and stops the patch floating on an invented white void.

Open the result full size and look for the details that make it your subject rather than a generic one — a collar tag, a scar, a logo-free badge. If they vanished, name them in the prompt and rerun.
Four jobs, and none of them is a photo filter for its own sake.
Showing a client the patch on the cap is a far easier sell than showing them the photo and describing the patch. It costs minutes rather than a sampling run.
A stitched dog reads as a keepsake in a way a printed one does not. Same instinct as a painted pet portrait, different material and a much shorter turnaround.
A flat logo rendered as a patch on denim carries workwear associations no vector file does. Useful for a lookbook page or a pitch deck cover.
Not a machine file, but a picture of the target. Handing an embroiderer an image of the finished patch you want removes an entire round of misunderstanding about fill direction and colour count.
What people ask before running a photo through thread.
Not directly. What you get is an image, and machines need a stitch file — .dst, .pes, .exp — with a stitch order and colour changes in it. A digitiser produces that from artwork. The image is a very good brief for them, because it settles what the finished patch should look like before anyone starts.
One clear subject, a plain background and even light. Pets, faces, single objects and logos all convert well. Group shots and busy scenes convert badly, because embroidery simplifies aggressively and a crowded frame simplifies into mush.
Broadly yes, and that is the thing to check every time. Markings, pose and colour carry through; very fine detail does not, exactly as it would not in real stitching. If a defining detail disappears, name it explicitly in the prompt and run again.
Yes. Describe the palette you want — three colours, high contrast, a cream border — and the conversion follows it. Restricting the count usually looks more authentic than letting it sample the whole photograph, because real patches are limited by how many thread changes are worth making.
On bold, simple marks, yes. Fine lettering is where it struggles, for the same reason it struggles in the real craft: small type has to be stitched larger than people expect. Keep any wording short and heavy if it matters.
It travels over HTTPS, goes into your own history, and nowhere else. It is never used to advertise, and deleting it removes it. Every output is AI-generated.
One credit pool covers Nano Banana images and Seedance and Veo video. The cost shows before every run, and failed jobs are not charged. Use a subscription for ongoing work, or a one-time pack when you just need to top up.
What's included
Secure checkout by Stripe. Card details never touch our servers.
What's included
Secure checkout by Stripe. Card details never touch our servers.
What's included
Secure checkout by Stripe. Card details never touch our servers.
What's included
Secure checkout by Stripe. Card details never touch our servers.
One-time top-ups — buy extra credits any time you run low.
What's included
Secure checkout by Stripe. Card details never touch our servers.
What's included
Secure checkout by Stripe. Card details never touch our servers.
What's included
Secure checkout by Stripe. Card details never touch our servers.
Charges appear as “KAVEL AI” on your card statement. You can cancel any time from Settings → Billing; cancellation takes effect at the end of the period you already paid for. Operator details, the full model list, and the refund window are on the about page.
Powered by
Use one balance across every supported image and video model — all of them listed below — and check the credit cost before each request.
Video models
Image models
Model credit guide
You're only charged for successful generations. The exact estimate in the generator varies by model, length, resolution, audio, and number of images.
| Type | Model | Credit cost |
|---|---|---|
| Video | Seedance 2.5 | The long-take tier, priced per second: 5s 480p ≈ 211 credits, 5s 720p ≈ 473. A full 30s take runs ≈ 1,260 at 480p and ≈ 2,835 at 720p. Supplying a reference clip lowers the per-second rate. |
| Video | Seedance 2.0 | 5s 720p image-to-video ≈ 188 credits; text-to-video ≈ 308 credits. Scales with resolution and length. |
| Video | Seedance 2 Fast | Faster and lower cost. 5s 720p text-to-video ≈ 248 credits. |
| Video | Seedance 2 Mini | The cheapest Seedance tier. 5s 720p ≈ 154 credits, 5s 480p ≈ 72 credits. |
| Video | Seedance 1.5 Pro | Audio doubles the rate. 5s 720p ≈ 27 credits silent, ≈ 53 with audio; 1080p ≈ 57 and ≈ 113. |
| Video | Veo 3.1 | Billed per video, not per second (Lite tier). About 45 credits at 720p and 53 at 1080p. |
| Video | Kling 3.0 | Audio raises the rate. 5s 720p ≈ 105 credits silent, ≈ 150 with audio; 1080p ≈ 135 and ≈ 203. |
| Video | MiniMax H3 | Fixed 2K, no resolution ladder. Priced per second — a 5s clip ≈ 158 credits. |
| Image | Nano Banana 2 | Generate or edit from text and images. 20 credits per 1K image, 30 at 2K, 45 at 4K. |
| Image | Nano Banana Pro | Consistent run times across generations. 30 credits per 1K or 2K image, 50 at 4K. |
| Image | Nano Banana 2 Lite | Faster, simpler variant, and the everyday editing price: a flat 8 credits per image at every resolution. |
| Image | GPT Image 2 | The lowest-cost premium image model here. 10 credits per 1K image, 15 at 2K, 30 at 4K. |
| Image | Seedream 5.0 Lite | Flat pricing — 20 credits per image whether you generate or edit. |
Upload one clean shot, name the fabric behind it, and check the stitch direction at full size — or try the same subject in clay.