Turn Any Photo Into Art With AI
By Chatday Editorial Team ·
A while back, a lot of people turned their selfies into little collectible figurines. They gave a photo to an AI, asked for a toy version of themselves, and posted the result. It got so popular that Google’s image tool reportedly gained about 13 million new users in four days, and people made hundreds of millions of images in the first two weeks.
Here is the part most people missed: that trend was not a special feature. It was just one style. The same trick turns your photo into a watercolor, a pencil sketch, a 90s anime frame, a Renaissance oil painting or a claymation still. You upload a picture, you describe the look you want, and the AI repaints it.
This guide shows you how to do it well, not just once by luck. What kind of photo to start with, how to write the style request, which look suits which photo, how to fix a result that comes out wrong, and the honest limits you should know before you post anything.
What “turning a photo into art” actually means
There are two ways an AI can make a picture. It can invent one from a text description, or it can take an image you already have and change it. The second one is what we want here, and it is called image-to-image, or image editing.
The difference matters. When you generate from scratch, you get a stranger who looks a bit like your idea. When you edit your own photo, the AI keeps the real shapes, the real pose and the real composition, and only swaps the surface: the lines, the colors, the texture, the light. That is why a restyled photo of your dog still looks like your dog, just painted.
So the whole job is two decisions. Pick a good photo to start from, and describe the style clearly. Get those right and the rest is easy.
Start with the right photo
The AI is working from what it can see. A sharp, well lit photo gives it plenty to keep. A dark, blurry one gives it very little, so it starts inventing, and that is when faces come out distorted.
- Good light beats everything. Daylight near a window is perfect. Avoid harsh shadows across a face.
- One clear subject. A busy background full of people and objects confuses the result. Simpler is better.
- Get close enough. The face, the pet or the object should fill a decent part of the frame.
- In focus. If a detail is blurry to you, it is blurry to the AI, and it will guess.
Step 1: Name the style properly
“Make this artistic” gives you a generic, bland result. The trick is to describe three things: the medium (what it is made of), the style (the specific look), and the mood (the feeling). Upload your photo, then use a prompt like these.
Notice the pattern. Each one says what the picture is made of, what look it should have, and then a firm instruction to keep the real subject. That last line is what stops the AI from redrawing a whole new person.
A quick guide: which style suits which photo
Not every look works on every photo. Here is a rough map to save you time, based on what tends to hold up and what tends to fall apart.
| Your photo | Styles that work best | Styles to avoid |
|---|---|---|
| Close-up portrait | Watercolor, oil painting, pencil sketch, pop art | Busy 3D scenes (faces distort) |
| Full-body photo | Cartoon, toy figurine, comic book, claymation | Fine detail styles (loses the face) |
| Pet photo | Watercolor, cartoon, storybook illustration | Photoreal restyles (uncanny) |
| Landscape or travel | Watercolor, ink wash, vintage poster, oil | Portrait-focused styles |
| Group photo | Poster, comic panel, flat illustration | Anything needing sharp faces |
The pattern behind it: the more faces and fine detail a photo has, the simpler and flatter the style should be. Painterly and illustrated looks forgive small errors. Hyper-detailed 3D looks expose them.
Step 2: Fix a result that comes out wrong
Your first try is a draft, not the final. The best part of doing this with a chat-style tool is that you can just ask for changes in plain words, and it edits what it already made instead of starting over.
- Face changed? Say: “same image, but keep the original face and expression exactly.”
- Too intense? Say: “same style, but softer and more subtle, less saturated.”
- Lost a detail you liked? Name it: “keep the red scarf and the glasses.”
- Want a variation? Say: “same photo, now as a pencil sketch instead.”
Small, specific corrections beat rewriting the whole prompt. You are nudging, not restarting.
Best styles for each kind of photo
If you want specifics, here are the looks that reliably get the most “how did you make that” reactions.
- Best for a profile picture: oil painting or watercolor portrait. Flattering, timeless, and kind on a day you do not love how you look.
- Best for a gift: a storybook or watercolor version of a pet or a family photo, printed and framed.
- Best for social media: the toy-figurine or cartoon style, because it reads instantly at thumbnail size.
- Best for a travel post: turn the landscape into a vintage poster or an ink wash. It looks designed, not just filtered.
- Best for a group photo: a flat comic-panel style, which turns an awkward group shot into something that looks intentional.
Where AI restyling falls short
This is genuinely fun, but it is not magic, and knowing the weak spots saves you from posting something odd.
Faces and hands. These are still the hardest thing for AI. On a group photo especially, a face in the background can come out melted, and hands sometimes gain or lose a finger. Zoom in and check before you share.
Text. Any writing in the photo, a sign, a logo, a t-shirt slogan, usually turns into nonsense letters. If the words matter, this is not the tool for it yet.
It can lose the likeness. Push the style too hard and the person stops looking like themselves. If keeping the resemblance matters, ask for a lighter touch and keep the “keep the same face” instruction in every prompt.
Living artists and brands. Asking for the exact style of a specific living artist is an ethical grey area, and copying a brand’s characters can cross into their rights. Describe the look you want (soft watercolor, bold comic ink) rather than naming a person to copy. You will get a better, cleaner result anyway.
Every photo is a starting point
That is really the whole idea. Every photo on your phone is a starting point, not a finished thing. The birthday snapshot can become a watercolor. The dog can become a storybook illustration. The holiday view can become a poster you would actually hang.
Pick one photo with good light, name the style you want in plain words, and let the AI repaint it. If the first try misses, nudge it. Ten minutes from now you will have something worth framing, and probably three more you did not expect to love.