Upload a photo and get it back as a warm, hand-painted anime scene: soft cel-shading, gentle light and rounded linework, the look popularized by Japanese animation studios. Compare a dedicated photo-to-anime model against general editors, then iterate the result until the mood matches the reference in your head.
Studio Ghibli films share a visual signature almost anyone recognizes on sight: soft cel-shaded light, rounded character design, skies painted in layered watercolor clouds, and a warmth that reads as hand-drawn rather than rendered. That signature turns out to be a very learnable style for an image model, which is why photo-to-anime conversion has become one of the most requested edits on Picasso IA. This page walks through the real models that do it, what to type, and where the illusion breaks.
A photo carries information a prompt alone cannot specify: exact pose, precise lighting, the particular geometry of a face or a street. Feeding that photo into an editing model as a reference solves the hardest part of the problem, the model only has to reinterpret what already exists rather than invent it from text. That is a narrower job than generating a scene from scratch, and narrower jobs are where current image models are most reliable.
The Ghibli look specifically rewards this approach because it is defined by consistent, describable choices: cel-shaded flat color instead of photographic gradients, soft rounded outlines, warm natural light, and skies that look painted rather than photographed. Models trained on large volumes of anime and illustration data have absorbed those choices well enough to apply them to an arbitrary photo convincingly, which is a very different task from drawing an original Ghibli-style character from nothing. One honest note before you start: Picasso IA is not affiliated with, and does not claim any license from, Studio Ghibli. What ships here is a style description an image model can follow, the same way "watercolor" or "film noir" is a style, not a trademarked filter.
Picasso IA runs 488 models behind one interface, and a handful are built or well suited for exactly this restyle. They behave differently enough to be worth trying more than one on the same photo before you settle on a result; the full catalog is browsable if you want to see what else is in there.
| Model | Best for | What to type |
|---|---|---|
| qwen-image-edit-plus-lora-photo-to-anime | A dedicated, one-step cel-shaded conversion | transform into anime, clean line art, vibrant cel-shaded colors |
| nano-banana-2 | A painterly, warm-toned restyle that keeps the original mood | Studio-Ghibli-inspired anime illustration, soft cel-shading, warm colors, dreamy atmosphere |
| flux-kontext-pro | A targeted style edit that leaves composition untouched | using this style, redraw the photo as a soft painterly anime scene |
| cartoonify | A flatter, simpler cartoon look when full anime detail is too busy | no prompt needed, one click restyles the upload |
| qwen-image-edit-plus | Combining the restyle with a second change in the same pass | your restyle instruction plus a second edit, such as a background swap |
The photo-to-anime LoRA is purpose-built for this and needs the least prompting, the fastest starting point. Nano banana 2 and Flux Kontext Pro are general editors that happen to be good at style transfer when the prompt is specific about mood and shading, worth a second pass when the dedicated model gets the linework right but the light feels off.
This is the workflow that gets from an ordinary photo to a usable illustration without a dozen wasted generations. It runs entirely in the Picasso IA toolkit, and new accounts include free credits, enough to try it on a few photos before spending anything.
The words you add on top of "anime" or "Ghibli-style" are what separate a generic cartoon from something that actually reads as hand-painted. These five habits do most of the work.
Pro tip: restyle a face-forward portrait before you try a group shot or a full scene. A single clear subject shows you exactly what a model does to eyes, hair and skin tone in this style, and that result tells you more about what to expect than three attempts at something more complicated.
The obvious use is a personal portrait turned into a keepsake illustration, a profile picture, or a print for a gift, and that alone accounts for most requests. Beyond that, small businesses use the same restyle for warm, illustrated social posts that stand out from photography in a feed, and independent authors use it for book covers and chapter art when a painterly look fits the story better than a photograph would.
Anyone building a recurring character in this style runs into the same identity problem every restyle workflow faces: the model reinterprets a face slightly differently each time. The fix is the same technique used for consistent AI characters, lock a reference image of the restyled result and feed it back in for the next scene instead of restyling the original photo again from scratch. Portrait work specifically benefits from starting with a clean, well-lit source; the guidance in AI headshots on lighting and framing applies just as well here, since a better input photo produces a better restyle. For a broader look at how a multi-model catalog compares to a single style-focused tool, the Picasso IA vs Nano Banana breakdown covers where each approach wins.
The restyle is convincing at a glance and imperfect up close. Hands and complex poses are the weakest spot: a restyled hand can gain or lose a finger, and an unusual pose sometimes bends in a way a hand-drawn artist never would, so check hands specifically before you use an image publicly. Text and fine background detail hold up next worst, since signage, small patterns and busy backgrounds often come back smudged or simplified as the model spends its attention on the main subject instead.
Group photos are harder than single portraits for the same reason identity is harder in general: every additional face in the frame is one more chance for the model to drift from the source photo, so a family of five restyles less predictably than one person. And this produces a style inspired by a well-known aesthetic, not a specific film frame or an official filter; results vary between models and even between runs of the same model, so treat consistency across a batch as something you check rather than assume.
None of this makes the restyle unusable, it means treating the first result as a draft. Run a second pass on anything that looks off before you call it finished.
qwen-image-edit-plus-lora-photo-to-anime is the fastest starting point because it was built specifically for cel-shaded anime conversion and needs almost no prompting. Nano banana 2 and flux-kontext-pro are worth a second pass when you want more control over mood, lighting or a detail the dedicated model got wrong. Running the same photo through two of the 488 models and keeping whichever result feels right is normal, not a sign something failed.
Picasso IA makes no claim to any Studio Ghibli license or affiliation; what the tool applies is a describable visual style, similar to how "watercolor" or "film noir" describe a look rather than a trademark. Personal use, like a keepsake portrait or a gift, is generally uncontroversial. If you plan to sell or publish restyled work commercially, that is a legal question that depends on your country and how the image is used, and it is worth checking rather than assuming.
Some reinterpretation is unavoidable, since the model is repainting the image rather than filtering it, but a clear, well-lit, front-facing photo keeps recognizable features like glasses, hairstyle and expression largely intact. If a first result feels off, naming the specific feature you want preserved in the prompt, rather than rewriting the whole request, usually fixes it faster.
Yes, and pets and landscapes often restyle more cleanly than people, since there is no face to reinterpret and no risk of a hand looking wrong. Landscapes benefit especially from the painted-sky and warm-light prompt words, since skies are where the Ghibli look reads most clearly.
A generic cartoon filter, like cartoonify, applies flat color and simplified lines quickly and works well for a fast, playful result. The photo-to-anime and painterly-editor route described here specifically targets cel-shading, warm natural light and painted backgrounds, the details that make an image read as hand-illustrated rather than just simplified. Try both on the same photo if you are unsure which look you want.
No. Every model described here takes a plain-language instruction, not a design tool. Upload a photo, describe the style you want in a sentence, and adjust the wording if the first result misses the mood. The effects page also collects preset looks if you want a starting prompt rather than writing one from scratch.
Yes, but expect more variation than with a single subject, since each additional face is another chance for identity to drift during the restyle. Running a group photo through the dedicated photo-to-anime model first and reviewing which faces held up best is the most reliable approach; a second pass with a general editor can then fix individual faces that came out wrong.
Generations are paid in credits, and the cost depends on the model and settings you choose, so a specific number here would be wrong within a month. Image restyles sit at the cheap end of the catalog compared to video or 3D generation. New accounts include free credits, and current plan pricing lives on the pricing page, which is the only place worth trusting for numbers.
Yes. Run the finished restyle through an upscaling model before exporting, which sharpens fine linework that can look soft straight out of the initial generation. This matters most for a large print, where any softness in the original becomes obvious.
Yes, qwen-image-edit-plus and flux-kontext-pro both accept one prompt describing the style change and a second change together, such as restyling a portrait and swapping the background in one pass. Doing changes one at a time across two generations usually gives more control than asking for everything at once.
Your photo is one upload away from looking hand-painted. Open the Picasso IA toolkit and see what the photo-to-anime model does with it.
