AI Style Transfer
Apply any art style to your photo while keeping the subject recognisable. Oil painting, watercolour, anime, 3D render and more. Free credits on signup.
What can you create with AI Style Transfer?
AI style transfer from photo
Upload a photo, name the style, and the subject survives the transformation. That last part is the hard bit: plenty of tools will happily turn your friend into a beautiful oil painting of somebody else. Describing the style in terms of medium and era — 'thick impasto oil, late Van Gogh' rather than just 'painting' — gives the model something concrete to hold onto and keeps the likeness intact.
Styles you can apply
Oil, watercolour, gouache, ink wash, anime cel, comic halftone, 3D render, low-poly, papercraft and pencil all work reliably. Photographic looks work too: cross-processed film, tungsten-lit reportage, large-format portrait. The styles that struggle are the ones defined by a single artist's living work — a style is a technique, and the more you describe it as technique rather than as a name, the better the result.
Keeping faces recognisable through a restyle
Faces are where style transfer usually falls apart. Two things help. Start from a well-lit photo where the face occupies a decent share of the frame, and pick a style that keeps facial structure — anime and 3D render preserve likeness better than heavy impasto or abstraction. If a face drifts, running the same photo at a lighter stylistic intensity almost always recovers it.
Style transfer vs a plain filter
A filter maps colours. Style transfer redraws the image. The difference shows up in texture: a watercolour filter tints your photo, whereas a watercolour restyle gives you visible paper grain, bleeding edges, and brush behaviour that follows the forms in the picture. It also shows up in what gets simplified — a real restyle drops detail the way the medium would, instead of keeping every pixel and recolouring it.
Best model for style transfer
Flux Kontext is the specialist here; it was built for instruction-driven editing and holds the original composition while changing how it is rendered. Nano Banana 2 is the fast alternative and handles conversational follow-ups well, so you can say 'more contrast, less saturation' and iterate. Both are in the model picker on the same screen, so trying the other one costs a click.
What usually goes wrong, and how to avoid it
The two common failures are losing the person and losing the photo. Losing the person happens when the style is too aggressive for the face size in frame — crop tighter or pick a flatter style. Losing the photo happens when the model reinterprets the scene instead of restyling it, which is a sign the prompt described a new image rather than a transformation. Anchor it with 'keep the composition and subject; change only the rendering'.
How to use AI Style Transfer
- 1Sign in (free)
Create a SoonMaker account in one click with Google. You'll get free credits immediately — no card required.
- 2Upload a clear photo
Pick a well-lit photo where the face or subject is clearly visible and facing the camera — that gives the AI the most to work with.
- 3Apply the style and download
Run AI Style Transfer, compare the result with your original, and download it. Regenerate if you want a slightly different interpretation.