ThatPainter is reader-supported. When you buy through links on our site, we may earn an affiliate commission. Learn More

Yes—you can use ChatGPT to create an original hand-painted fantasy illustration or transform an uploaded selfie, pet photo, landscape, or meme into an animated image with a Studio Ghibli-like mood. The current workflow uses ChatGPT Images 2.0, which OpenAI currently documents as available on the web, iOS, and Android across all ChatGPT tiers. Free accounts can generate images, but image creation has its own usage limit and is not unlimited.

For the most consistent results, do not rely only on the words “Studio Ghibli style.” Describe the visual ingredients you want—painted backgrounds, soft natural light, expressive faces, lush environmental detail, watercolor texture, a muted palette, and a peaceful sense of wonder. Then refine the result with ChatGPT’s editor and selection tool.

How to Use ChatGPT to Generate Studio Ghibli-Style Images

As of August 9, 2026, you do not need a separate “Ghibli AI” website or a manual GPT-4o selection to make this kind of image. Open ChatGPT, choose Images from the sidebar—or ask for an image in a normal conversation—then either describe an original scene or upload an image to transform.

The practical process is:

1. Open ChatGPT on the web or in the iOS or Android app.
2. Choose Images, or start a normal chat and ask, “Create an image of…”
3. Describe the subject, composition, lighting, palette, texture, mood, and format.
4. Optionally upload a photo and state what must remain unchanged.
5. Generate the image.
6. Open the result and make focused edits with chat or the Select tool.
7. Change the aspect ratio if necessary, then use the image controls to Save or Share it.

“Studio Ghibli-style” is a convenient user-facing shorthand, not an official Studio Ghibli filter, preset, endorsement, or guarantee that the result will resemble a particular film. A descriptive prompt is usually more controllable—and more clearly original—than a one-line request containing only a studio name.

What you need before you start

• A ChatGPT account and access to ChatGPT on the web, iOS, or Android.
• A source image, if you want to transform a selfie, pet photo, travel picture, or other existing image.
• Permission to use and transform any image you upload, including permission from people whose likenesses appear in it.
• A clear idea of the subject, mood, composition, and output shape you want.

OpenAI’s current documentation says ChatGPT Images 2.0 is available on all tiers, including Free. That does not mean unlimited generation: Free users have a separate image-generation limit, which may be stricter than the ordinary chat limit. If you reach it, ChatGPT displays the applicable reset time. Paid plans generally provide higher usage limits, but limits and availability can change. See OpenAI’s ChatGPT Images guide and current usage-limit guidance for the account-specific details.

For an uploaded image, OpenAI currently lists a 20 MB limit for individual images uploaded to ChatGPT conversations. A well-lit image with a clear subject generally gives the model more useful information than a dark, heavily compressed, or cluttered image. The upload limit is documented in the File Uploads FAQ.

Method 1: Transform a selfie, pet photo, or landscape

This is the quickest route when you want to preserve a real person, animal, building, or location. ChatGPT can reinterpret an uploaded image conversationally and can also make later changes with its image editor.

1. Open ChatGPT and select Images from the sidebar, or open a normal conversation.
2. Use the attachment control to upload your photo.
3. Paste a transformation prompt. Say which parts of the image must remain recognizable.
4. Generate the result. Complex image requests can take a few minutes.
5. Select the generated image to open its editor.
6. Use a direct follow-up instruction or select a specific area before requesting a local edit.
7. Use Aspect ratio if you need a square, portrait, landscape, or wallpaper crop.
8. Choose Save to download the image or Share to send it elsewhere.

Photo-transformation prompt

Reinterpret the uploaded photo as an original hand-painted Japanese animated-film illustration.
Preserve the person’s identity, pose, clothing, camera framing, and the main background layout.
Use soft watercolor texture, warm natural lighting, expressive but natural facial features,
detailed environmental painting, gentle atmospheric perspective, and a whimsical,
peaceful mood. Do not add logos, captions, recognizable copyrighted characters, or extra people.
Portrait format, 4:5.

The identity and preservation instructions matter because a broad “make this Ghibli style” request may change more than the rendering technique. Depending on the source image and the requested edit, ChatGPT may alter facial proportions, clothing, background objects, or the person’s apparent age. “Preserve the pose and clothing” reduces those changes but cannot guarantee pixel-level fidelity.

If you want to use the studio name

You can test a prompt that includes the name, but understand that it may behave differently from a descriptive prompt. ChatGPT’s systems can rewrite, refuse, or alter requests, and OpenAI says its image system checks text prompts, uploaded images, and generated outputs.

Transform this photo into a Studio Ghibli-inspired animated illustration while preserving
 the person’s identity, pose, clothing, and composition. Use soft painted backgrounds, warm
 natural light, expressive faces, lush environmental detail, and a quiet whimsical atmosphere.
 Do not add logos, film characters, or text. Portrait format, 4:5.

Do not treat descriptive wording as a way to evade a refusal. It is useful because it gives the model concrete artistic direction and encourages an original scene. OpenAI’s ChatGPT Images 2.0 safety overview explains that prompts, inputs, and outputs can be subject to safety checks, while OpenAI’s Terms of Use prohibit bypassing safety mitigations.

Method 2: Generate an original scene from text

Text-only generation is better when you want an original character, fantasy landscape, rural train station, cottage, seaside village, forest, wallpaper, or cinematic scene. It avoids the need to upload a personal photograph and gives you freedom to control the setting.

1. Open ChatGPT and choose Images, or ask for an image in a normal conversation.
2. Describe the subject and what is happening in the scene.
3. Add the camera position or composition, environment, lighting, colors, texture, and mood.
4. State that the character and design should be original if you do not want recognizable franchise elements.
5. Specify the aspect ratio or intended use.
6. Generate the image, then revise specific problems rather than starting from scratch each time.

Original-scene prompt

Create an original hand-painted Japanese fantasy-animation illustration of a young traveler
standing beside a quiet rural train stop at sunset. Use soft watercolor-and-cel-painted textures,
warm amber light, expressive but natural facial features, lush grass and distant mountains,
subtle atmospheric perspective, gentle wind moving the clothing, and a peaceful, nostalgic mood.
Use an original character design with no recognizable film characters, logos, or text.
Landscape composition, 16:9.

This prompt separates the request into controllable parts: subject, setting, medium, light, atmosphere, originality constraints, and format. That is more reliable than asking for a named style with no information about what should appear in the image.

How to write a better prompt

A useful structure is:

[action] + [subject] + [composition] + [setting] + [lighting] +
[color palette] + [texture or medium] + [mood] + [constraints] + [format]

1. Describe the subject

Identify the central subject and the important details ChatGPT should prioritize:

• “A young traveler with a yellow raincoat and a red satchel.”
• “A small black dog with a white patch on its chest.”
• “A weathered cottage surrounded by tall grass.”
• “Two children walking through a moss-covered forest.”

For an uploaded photo, identify what must remain recognizable: the person’s identity, the dog’s markings, the building’s shape, or the landscape’s main layout.

2. Control the composition

Composition determines how the subject is framed. Include terms such as:

• “Close-up portrait” or “full-body portrait.”
• “Wide cinematic shot.”
• “Viewed from behind at child height.”
• “Centered subject with open space above.”
• “Overhead view.”
• “Keep the original camera angle and crop.”

3. Specify the environment

Concrete locations produce more useful results than a general request for a “nice background.” Try a cedar forest after rain, a quiet seaside village, a rural train stop, a mountain meadow, or a sunlit kitchen with plants on the windowsill.

4. Describe light, palette, and texture

For the soft, nostalgic animated-film mood many readers mean by “Ghibli-like,” describe the ingredients rather than relying on the label:

• Soft morning light, overcast daylight, golden hour, candlelight, or moonlight.
• Muted greens and ochres, warm amber, pale blue, or a natural low-saturation palette.
• Hand-painted backgrounds, watercolor washes, delicate cel-painted shading, and visible environmental detail.
• Detailed clouds, plants, distant scenery, atmospheric perspective, and gentle wind.

If you want to avoid an unwanted look, say so explicitly: “Avoid glossy 3D rendering, photorealism, neon cyberpunk colors, and harsh high-contrast lighting.” Negative instructions are not perfect, but they can make the intended direction clearer.

5. State the mood

Use a small number of emotional descriptors: peaceful, nostalgic, mysterious, playful, melancholic, adventurous, or quietly magical. “Whimsical but grounded” is useful when you want fantasy elements without a chaotic or overly cartoonish result.

6. Add preservation and exclusion constraints

For a photo, include specific constraints such as:

• “Preserve the person’s identity, age, hairstyle, pose, and expression.”
• “Keep the same clothing and camera angle.”
• “Keep the dog’s markings and body proportions.”
• “Do not replace the background.”
• “Do not add people, logos, captions, or accessories.”

For original work, use: “Create an original character design with no recognizable film characters, logos, or copied poster layout.”

7. Choose the aspect ratio

State the intended shape in the prompt, then use the editor’s aspect-ratio control when available:

Format	Useful for
1:1	Profile images and square social posts
4:5	Portrait-oriented social posts
16:9	Desktop wallpapers, banners, and cinematic landscapes
9:16	Phone wallpapers, Stories, and vertical video backgrounds

OpenAI documents aspect-ratio regeneration in the current image editor. Regenerating the image in a new shape may change the composition, so review the edges and subject placement afterward.

Prompt examples for common projects

Pet portrait

Turn the uploaded dog photo into an original hand-painted animated illustration.
Preserve the dog’s exact coat pattern, face markings, ear shape, and sitting pose.
Place the dog in a lush summer meadow with tall grass and small wildflowers.
Use warm late-afternoon light, soft watercolor texture, subtle cel shading,
a natural muted palette, and a peaceful whimsical mood. No text, collar logo, or extra animals.
Square format, 1:1.

Travel memory

Reinterpret this uploaded travel photograph as an original painted fantasy-animation scene.
Keep the main building, street layout, people’s positions, and perspective recognizable.
Use soft overcast light, hand-painted walls and foliage, delicate watercolor texture,
quiet environmental storytelling, and warm natural colors. Remove all signs and logos,
and do not add new people. Landscape format, 16:9.

Original forest scene

Create a wide cinematic illustration of two original children walking through a moss-covered
forest after rain, viewed from behind at child height. Use soft painted backgrounds, delicate
watercolor texture, warm green and ochre tones, shafts of diffuse sunlight, expressive
environmental detail, and a quiet sense of wonder. Original characters only, no text or logos.
16:9 landscape.

Portrait avatar

Transform the uploaded portrait into an original hand-painted animated avatar.
Preserve the person’s identity, hairstyle, facial proportions, expression, and clothing.
Use a softly painted background with warm morning light, restrained cel shading,
natural skin tones, expressive but realistic eyes, and a calm optimistic mood.
No jewelry added, no text, no logos. Portrait format, 4:5.

How to refine the first result

The first generation is a starting point, not necessarily the finished image. The current editor supports direct text instructions, selection-based changes, aspect-ratio regeneration, undo, and redo. OpenAI cautions that selections are not always precise and edits can extend beyond the highlighted area.

Use one change per turn

Make a controlled sequence of revisions:

1. Correct the face or identity.
2. Fix the background.
3. Adjust the lighting.
4. Correct clothing or object details.
5. Change the crop or aspect ratio.
6. Add or correct text last.

Changing everything at once makes it difficult to determine which instruction caused an unwanted alteration. If a later edit makes the image worse, use Undo or return to the best earlier version and branch from there.

Use the Select tool for local edits

Select the relevant area, then describe only the intended change. Name the area in the instruction even when it is highlighted:

Replace only the background with a misty cedar forest. Keep the person, pose, clothing,
lighting direction, and facial expression unchanged.

Change only the sky to a soft pink-and-gold sunset. Preserve every foreground object.

Remove the extra character on the left and reconstruct the background naturally.

Keep the composition unchanged, but make the colors slightly softer and the background
more painterly.

Selection boundaries can spill over, particularly around hair, hands, transparent objects, and detailed edges. If an edit changes too much, select a smaller area and make the instruction narrower.

Useful correction prompts

Keep the face closer to the uploaded reference. Preserve the person’s age, hairstyle,
facial proportions, expression, and clothing. Change nothing else.

Correct the hands and facial proportions only. Keep the same person, clothing, pose,
lighting, background, and overall composition.

Change only the lighting and background texture. Preserve the person, pose, clothing,
face, camera angle, and all major objects exactly.

What to do when the image looks like generic anime

“Generic anime” usually means the prompt did not provide enough information about the medium, environment, lighting, or emotional tone. Add concrete visual direction:

Use a hand-painted environmental background, soft natural light, muted greens and ochres,
subtle watercolor texture, expressive but restrained facial features, and detailed clouds,
plants, and distant scenery. Avoid glossy 3D rendering and neon cyberpunk colors.

Also check the composition. A prompt that specifies only “a Ghibli-style character” leaves the setting, camera angle, palette, and atmosphere underdetermined. The more important those choices are to your image, the more explicitly you should state them.

Using film references, characters, and logos

A broad visual description is not the same request as copying a specific work. These requests carry different creative and rights considerations:

Request	What it means	Practical consideration
Style-inspired original artwork	An original character or scene using broad visual traits	Usually gives you the most control and the least dependence on a protected franchise
Photo transformation	An uploaded image reinterpreted as animation	Requires image rights and appropriate consent; identity may still change
Fan art	A recognizable copyrighted character or fictional world	May involve copyright, trademark, or other rights; publication and commercial use need care
Exact recreation	A film frame, poster, logo, or existing artwork reproduced closely	Highest risk of copying protected expression or creating a misleading association

It is safer and more original to request “a quiet rural fantasy scene with a hand-painted natural environment” than to ask for a recreated film frame. Film titles can communicate mood quickly, but they can also pull in recognizable characters, props, settings, or compositions.

Do not claim that an image is official Studio Ghibli artwork, licensed artwork, or endorsed by Studio Ghibli unless that is independently verified. Studio Ghibli’s own notice about unauthorized reproductions and related English notice concern unauthorized reproductions and merchandise. They should not be overstated as a definitive ruling on every AI-generated style prompt.

Troubleshooting common problems

“I do not see Images”

• Confirm that you are signed into the intended ChatGPT account.
• Look for Images in the sidebar, or ask for image generation in a normal conversation.
• Update the iOS or Android app if you are using mobile.
• Check whether you have reached the separate image-generation limit.
• If ChatGPT shows a reset time, wait until that limit resets.

OpenAI’s documented workflow and availability can be checked in its Images in ChatGPT help article. The separate image limit is described in its usage-limit FAQ.

The result changes the person too much

Use a focused follow-up instead of regenerating the whole concept:

Preserve the person’s identity, age, hairstyle, facial proportions, expression, clothing,
pose, and camera angle. Change only the background texture and lighting.

For the most important face or clothing correction, select only the affected area where possible. Review every revision manually; generated images are not reliable substitutes for identity verification or a precise likeness.

The background changes unexpectedly

Keep the original background layout and all major objects. Change only the rendering style
and lighting. Do not add, remove, or rearrange buildings, people, trees, or vehicles.

If the change is local, select the background rather than the subject. Remember that the selection tool is not perfectly precise.

The face, hands, or fingers are wrong

Correct the hands and facial proportions only. Keep the same person, clothing, pose,
lighting, background, and overall composition.

When repeated edits make the image progressively worse, return to the best earlier version. It is often more effective to branch from that version than to keep correcting a damaged one.

Text in the image is misspelled

ChatGPT Images 2.0 is designed to handle more detailed text instructions than earlier image systems, but generated lettering still needs proofreading. For a poster, invitation, sign, or social graphic:

1. Use as little text as possible.
2. Put the exact wording in quotation marks.
3. Specify capitalization and line breaks.
4. Ask for a clean, readable layout.
5. Proofread every word in the result.
6. For exact typography, add the final text in a design application.

OpenAI documents improved text generation, not a guarantee of perfect spelling or professional typesetting.

ChatGPT refuses the request

Do not repeatedly rephrase the same blocked request in an attempt to defeat a safeguard. Instead, clarify the legitimate creative goal:

• Remove recognizable copyrighted characters or logos.
• Replace an exact film-frame recreation with an original composition.
• Use an original character design.
• Remove misleading real-world context.
• Make clear that the request is a non-deceptive artistic transformation.

OpenAI says its image system can block prompts, uploaded images, and final outputs. A refusal is not an invitation to disguise the same request.

Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Support on Ko-Fi

Can you sell or publish a ChatGPT-generated image?

OpenAI’s terms may permit you to use, reprint, sell, or merchandise ChatGPT output, but that permission does not automatically settle copyright, trademark, publicity-rights, licensing, or commercial-clearance questions.

OpenAI’s current US Terms of Use say that, as between you and OpenAI and to the extent permitted by law, you own the output. The same terms warn that outputs may not be unique and that another user may receive similar content. OpenAI’s Help Center also says users may reprint, sell, and merchandise output under the applicable terms and content policies, whether it was generated on a Free or paid plan. Review the current Terms of Use and OpenAI’s output-use guidance before relying on those permissions.

There are several separate questions:

• Contractual use: What OpenAI’s terms allow you to do with the output.
• Copyright ownership: Whether your jurisdiction recognizes copyright in the particular AI-assisted result.
• Human authorship: Whether your own creative contributions are sufficient for copyright protection.
• Third-party rights: Whether the image includes a recognizable character, logo, protected design, copied composition, or someone’s likeness.
• Commercial impression: Whether the image suggests that Studio Ghibli created, approved, or sponsored it.

The US Copyright Office says AI-assisted output may be copyrightable when a human author determines sufficient expressive elements, but merely supplying prompts is not enough by itself. Its guidance on human authorship and its discussion of AI, stylistic imitation, and market effects describe a fact-specific and evolving area.

A broad artistic style is not the same as copying a specific protected work. However, a generated image can still create copyright, trademark, publicity-rights, or unfair-competition concerns if it reproduces recognizable characters, logos, film frames, distinctive designs, or a misleading commercial association. Laws differ by country. For a product, book cover, advertisement, paid commission, or other important commercial use, obtain appropriate legal and rights advice rather than relying on the phrase “AI-generated.”

Be honest about how the image was made

Do not present an AI-generated image as entirely human-created, hand-painted, official Studio Ghibli artwork, or licensed artwork when that is not true. OpenAI’s terms prohibit representing output as human-generated when it was not. A simple disclosure such as “Created with AI assistance using ChatGPT Images” is appropriate when the audience could reasonably mistake the image for an original hand-drawn illustration or official studio work.

OpenAI says images generated with ChatGPT, Codex, and its API include C2PA metadata and SynthID watermarks. These are provenance signals—not proof of ownership, originality, legal clearance, or Studio Ghibli authorization. OpenAI’s C2PA and image-provenance guidance explains that its verification tool can detect supported signals, but a missing signal does not prove that an image was not generated by OpenAI because metadata or watermarks can be stripped or degraded.

Privacy and consent when uploading photos

Use only images you have the right to upload. Before transforming another person’s face, obtain appropriate consent—especially if you plan to publish, sell, advertise, or use the result in a way that could mislead people. Avoid uploading sensitive personal images unless you understand the account’s applicable data, retention, and privacy settings. A stylized output can still reveal or imitate a real person’s identity.

For consequential uses, do not treat a generated portrait as reliable evidence of someone’s appearance, identity, age, or behavior. Review the result yourself and consider whether the person would reasonably expect the transformation and its publication.

Is “Studio Ghibli-style” an official ChatGPT filter?

No. ChatGPT Images 2.0 is a general image-generation and editing system. “Studio Ghibli-style” is a popular description for a cluster of visual qualities, not an official Studio Ghibli control. The result may be painterly, anime-influenced, whimsical, or simply generic depending on the prompt, source image, safety checks, and revisions.

For a stronger original result, describe what you want to see: a hand-painted natural environment, watercolor-and-cel-painted texture, warm natural lighting, expressive but restrained faces, lush foliage, detailed skies, gentle magical realism, and a quiet coming-of-age atmosphere.

A practical checklist before you publish

• Did you use a photo you own or have permission to use?
• Did people shown in the source image consent to the transformation and publication?
• Does the image contain a recognizable copyrighted character, logo, film frame, or copied poster design?
• Could a viewer reasonably mistake it for official Studio Ghibli artwork or an endorsement?
• Did you check faces, hands, clothing, signs, and small text manually?
• Does the commercial use require additional clearance in your country?
• Have you disclosed AI assistance where the audience might otherwise be misled?
• Have you kept the original and best intermediate versions in case a later edit damages the result?

Frequently Asked Questions

Is ChatGPT Images free?

OpenAI currently documents ChatGPT Images 2.0 as available on all ChatGPT tiers, including Free. Free users have a separate image-generation limit, however, so free image creation is not unlimited. ChatGPT shows the applicable reset time when that limit is reached.

Do I need GPT-4o or a special model?

No manual GPT-4o selection is required for the current workflow. Use Images from the ChatGPT sidebar or request an image in a normal conversation. The separate DALL·E GPT remains a legacy route, not the default ChatGPT Images 2.0 instruction.

Can I use a selfie or pet photo?

Yes. Upload the image in a ChatGPT conversation and state which details must remain unchanged, such as identity, pose, clothing, markings, and camera angle. Use only images you have permission to upload, and obtain appropriate consent before publishing another person’s likeness.

Can ChatGPT create Totoro or another recognizable Ghibli character?

A request for a recognizable copyrighted character is different from an original image using broad visual traits. It may be refused or may raise copyright, trademark, licensing, or commercial-use concerns. For the lowest-risk original work, request an original character and do not include film logos, copied frames, or recognizable franchise designs.

Why did ChatGPT refuse my image prompt?

OpenAI says the current image system checks text prompts, uploaded images, and generated outputs. A refusal can result from the requested subject, likeness, copyrighted elements, misleading context, or another safety concern. Remove exact recreations, recognizable characters, logos, or deceptive context rather than trying to bypass the safeguard.

Can I sell a ChatGPT-generated image?

OpenAI’s terms and Help Center say output may be used, sold, or merchandised subject to the applicable terms and policies. That does not guarantee copyright protection or clear third-party rights. Review the image for characters, logos, copied designs, likeness rights, and misleading commercial associations; obtain legal advice for significant commercial projects.

How do I make the result less generic?

Specify the subject, camera angle, composition, environment, lighting, palette, medium, texture, mood, exclusions, and aspect ratio. For example, request hand-painted backgrounds, soft watercolor texture, muted greens and ochres, diffuse sunlight, detailed foliage, and restrained facial expressions instead of writing only “Ghibli style.”

How do I preserve a person’s face?

Tell ChatGPT to preserve the person’s identity, age, hairstyle, facial proportions, expression, clothing, pose, and camera angle. Use a clear source image and make focused edits with the selection tool. These instructions reduce unwanted changes but cannot guarantee an exact likeness.

What aspect ratio should I use?

Use 1:1 for a profile image or square post, 4:5 for a portrait social post, 16:9 for a cinematic landscape or desktop wallpaper, and 9:16 for a phone wallpaper or vertical Story. You can specify the format in the prompt and use the editor’s aspect-ratio control afterward.

Does a ChatGPT-generated image contain AI provenance information?

OpenAI says images generated with ChatGPT, Codex, and its API include C2PA metadata and SynthID watermarks. These signals communicate provenance; they do not prove ownership, originality, legality, or Studio Ghibli authorization. Signals can also be stripped or degraded.

The Bottom Line

To make a Studio Ghibli-inspired image with ChatGPT, open Images, describe an original scene or upload a permitted photo, and specify the composition, painted texture, lighting, palette, mood, constraints, and aspect ratio. Refine the result one change at a time with the editor and selection tool. ChatGPT Images 2.0 is currently available across all tiers, but Free users have separate usage limits.

For work you publish or sell, keep the result original, do not imply Studio Ghibli approval, check recognizable characters and logos, respect people’s likeness rights, and disclose AI assistance when readers could otherwise be misled.

Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.