One photo in · A finished painting out · Runs in the browser
Not a filter laid over a photograph. The washes bleed into each other, pigment gathers where the shadows were, and the brightest parts are simply paper the brush never touched.
Both frames come from the same source picture. Drag the divider to compare. Every output is AI-generated, so your finished piece varies with your photo, your prompt and the style you describe — read this as what the tool can do, not as a fixed result you will get back.
Drag to compare before and after
It repaints a photograph as a watercolour: transparent washes instead of pixels, soft bleeding boundaries instead of edges, and untouched paper where the highlights were. The subject stays recognisable; everything around it gets looser.
The old way to do this was a Photoshop action — a texture layer, a smudge pass, maybe a paper overlay multiplied on top. It always looked like a photograph wearing a costume, because the underlying image was still a photograph. What changed is that the model repaints the picture rather than processing it: it decides where a brush would have loaded pigment, where it would have run dry, and where to leave the sheet blank.
That distinction shows up first in the background. A filter has to keep every leaf behind your subject, because it has no way to decide which ones matter. A painter drops them into two or three washes of green and lets the shapes go. So does this. If you have already tried the oil painting portrait tool, this is the same idea with the opposite physics — that one builds up opaque paint, this one lets the paper show through.
Watercolour is also the most forgiving style to try on a picture you would not normally call good enough. Harsh phone flash, a busy background, a slightly soft focus — all of it becomes wash. The technique hides exactly the flaws a sharper style like the AI sketch generator would expose.
Creative engine
Upload the photo, say how loose you want the washes, pick the format and resolution — the credit estimate updates before you generate.
Colour is laid down as transparent pools that bleed where they meet, rather than as a texture multiplied over the photograph you started with.
The brightest areas are left unpainted instead of brightened, which is the single detail that separates a painting from a processed photo.
No brush settings and no layer stack. The only real decision is how loose you want it.
Anything with a clear main subject works — a person, a pet, a flower, a building. Backgrounds do not need to be tidy, because they are the first thing to dissolve into wash.
The preset asks for wet-on-wet bleeding and visible paper grain. Push it further with words like loose, minimal, or heavy bleed; pull it back with tight, controlled, botanical study.
One pass produces the finished piece at full resolution, ready to print or post.
The subjects that survive the technique best, in the order people reach for them.
Fur is the ideal watercolour subject: it is already soft-edged, so the wash does not have to fight it. A pet photographed on grass in full sun comes back with the lawn reduced to two washes and the animal intact. If you want the formal treatment instead — velvet, gold frame, a crown — that is the AI pet portrait generator.
Faces need the loosening kept away from the eyes and mouth, which is where the model puts its detail by default. The result is the sort of piece people frame — recognisable up close, painterly from across the room.
Invitations, menus and thank-you cards all want an image that sits behind type without competing with it. A wash does that; a photograph does not.
Landscapes and buildings lose their fussiness in watercolour — a bridge becomes its shape, a river becomes movement, and the sky becomes the paper.
Drop the pigment and keep the bleed and you get the monochrome variant: greys running down the page, one or two darks holding the composition together.
Clothing and background dissolve into splatter while the face holds photographic detail — the editorial look, without the illustration commission.
Yes, and that is where the model spends its detail budget. Features stay in the same places at the same proportions; what loosens is the clothing, the background and the outer edges — the same order a painter would work in.
Less than for most tools. Harsh light, a cluttered room and slightly soft focus all get absorbed into wash. The one thing that does matter is a clearly separable subject — if you cannot tell where the subject ends, neither can the model.
An action processes the pixels you already have, so every leaf and every hair stays where it was and gets textured. This repaints the picture, which means it can decide that forty leaves become three washes. That decision is the whole style.
Ask for it in the prompt — monochrome ink wash, greys only, minimal pigment. The bleeding behaviour is the same; you are only removing the colour.
The download is full resolution, which is enough for cards, prints and small framed pieces. For anything wall-sized, run it through the HD photo converter afterwards.
Credits are spent per generation, and the estimate is shown before you press the button. Paid credits open the stronger image models, which hold a likeness better once the washes get loose.
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. |
Different media, same upload — and the model doing the painting.
The bleed at the edges is the part a filter never gets right. It takes one upload to see whether this one does.
A real edit made with this tool — drag your own photo in on the left, the prompt is already written, then press Generate.
