I Turned a Photo of Me and My Wife Into a Japanese Manga — ChatGPT Got It Scarily Right
How I Turned a Photo Into a Japanese Illustration Using ChatGPT
This illustration didn't start with an AI-generated image. It started with a real photo of me and my wife during one of our favorite moments together. Out of curiosity, I uploaded the photo to ChatGPT and asked it to transform us into a Japanese manga-style illustration. A few prompts and refinements later, the result looked like something straight out of a slice-of-life romance manga. In this guide, I'll show you exactly how I did it so you can create your own.
A few years ago, creating a custom illustration like this would have required hiring an artist, waiting days for revisions, and spending a significant amount of money. Today, ChatGPT can transform a personal photo into stunning artwork in minutes.
The image featured in this tutorial is based on a real photograph of me and my wife sitting together in a park. Rather than creating something completely fictional, I wanted to preserve the emotion, expressions, and atmosphere of the original moment while transforming it into a Japanese illustration style.
What surprised me most wasn't the quality of the artwork—it was how accurately ChatGPT captured the feeling of the original photo. The smiles, the relaxed pose, the warmth between us, and even the overall mood remained intact while being reimagined as a hand-drawn manga illustration.
In this tutorial, I'll walk through the exact workflow I used, including the prompts, refinement techniques, and styling instructions that produced the final result. Whether you're creating a gift for your spouse, designing a unique social media profile image, celebrating an anniversary, or simply experimenting with AI art, you'll be able to follow along step-by-step.
The illustration shown throughout this article was generated from a real photo of me and my wife. No Photoshop was used. Everything—from the initial transformation to the final refinements—was created entirely inside ChatGPT using image generation and prompt engineering techniques.
Why This Illustration Feels Different
Most AI-generated portraits focus on realism. This one intentionally leans into storytelling. Instead of trying to create a perfect photographic replica, the goal was to capture a memory and present it as if it were a scene from a Japanese manga or Korean webtoon.
The floating hearts, hand-drawn stars, musical notes, soft watercolor accents, and sketch-style background all contribute to a feeling of nostalgia and warmth. The result feels less like a photograph and more like a memory brought to life through illustration.
That's the real power of modern AI image generation. It's not just about creating pictures anymore. It's about transforming personal moments into art.
🔍 Image Anatomy — What Makes This Style Work
Now you know what to ask for. Let's build the prompt.
What You Need Before Starting
- ChatGPT Plus subscription — 4 image generation is a Plus feature. Free tier has limited access. Worth the $20/month for this alone.
- A clear reference photo — full body or 3/4 shot works best. The AI uses it for pose, clothing, and general likeness. Good lighting in the original = better illustration.
- 10–20 minutes — first generation takes seconds. Refinement takes a few rounds. Plan for 3–5 attempts to get it exactly right.
- Patience on likeness — AI won't nail an exact face. It captures general vibe, skin tone, hair, and style. Think "recognizable portrait" not "perfect replica."
ChatGPT uses your uploaded images for context only — not for training (per OpenAI's current policy for Plus users). Still, avoid uploading photos of minors or images with identifiable location metadata if privacy is a concern. Read OpenAI's current data policy for specifics.
Step-by-Step Tutorial — Create Your Japanese Illustration
Pick a photo where both people are clearly visible, well-lit, and ideally sitting or in a relaxed natural pose. The more natural the pose, the better the illustration feels. Crop tightly — don't upload a wide landscape shot where the subjects are tiny. Aim for subjects taking up at least 40% of the frame.
Natural outdoor light beats indoor flash. Side-by-side poses work better than one person partially obscured. Clear clothing details help the AI render outfit texture correctly. Avoid heavy filters — the AI works from actual visual data, not filter effects.
Open ChatGPT and start a fresh conversation. Don't use an existing thread with lots of prior context — a clean session gives a cleaner slate. Select GPT-4o from the model dropdown (it has the best 4 integration for reference-based image generation).
Click the image upload icon (paperclip or image button in the chat bar), attach your photo, and then paste the base prompt below into the same message. Send both together — the image and the prompt in one go.
Transform this photo into a Japanese manga-style illustration. Art style: - Clean ink line art with varying stroke weight - Shojo manga meets Korean webtoon aesthetic - Selective watercolor fill — characters colored, background in pencil sketch only - Hand-drawn doodle overlays: small hearts, stars, and music notes floating around the couple - Ink splatter texture on the ground for tactile feel Characters: - Preserve skin tone, hair color, general face shape, and clothing from the photo - Both people smiling naturally, warm and relaxed expressions - Keep the sitting pose and physical positions from the reference Background: - Tall trees rendered as loose pencil/ink sketch (not painted) - Pine branch details at top - Silhouetted figures in the distance (sketched, not colored) - Grass and ground in soft green wash Overall mood: Warm, romantic, playful — like a scene from a slice-of-life manga Aspect ratio: 1:1 square No text or watermarks in the image
ChatGPT will generate an image within 15–30 seconds. Look at it critically — check the line art quality, the face expressions, whether the selective color approach came through, and if the doodle overlays (hearts, stars, music notes) appeared. Most first generations get the style right but need facial or compositional tweaks.
Don't start over — iterate. Type refinements directly as follow-up messages in the same conversation. ChatGPT remembers your photo and previous generations. Use the refinement prompts below depending on what needs fixing.
// If faces look off: Make the faces more expressive and warm — bigger natural smiles, softer eyes. Keep the manga style but make them look like real people, not chibi characters. // If background is too detailed or colored: The background should be pencil sketch only — no color fill on the trees or ground. Keep all color on the two characters only. // If doodle elements are missing: Add more floating doodle elements: small hand-drawn pink hearts, yellow and blue stars, and musical notes scattered around the couple. Make them look hand-drawn, not digital. // If the ink line weight is wrong: Make the outlines thicker and bolder on the characters — the line art should feel like it was drawn with a fine-tip ink pen. Thinner lines on the background elements. // If clothing details are lost: Restore the clothing details from the reference photo — the plaid shirt pattern, the dark t-shirt with text, the casual shorts. Keep them clearly visible in the illustration. // If you want more whitespace / cleaner composition: Simplify the background — reduce the number of tree trunks, pull them back slightly. Give more breathing room around the couple. Keep the composition feeling airy and light.
Once you have a generation that's close to what you want, right-click the image and save it, or use ChatGPT's download button. The image will be 1024×1024 PNG at minimum. If you need higher resolution for print, use an upscaling tool like Topaz Gigapixel or Adobe Firefly's upscaler on the saved file.
If you're using this as a gift, social post, or profile image, open the downloaded file in Canva. Add a name, date, or short caption using a font that matches the illustration feel — "Caveat" or "Patrick Hand" from Google Fonts work well for that hand-drawn aesthetic.
The prompt isn't magic. It's just very specific instructions in the language the AI understands. Vague gets generic. Specific gets art.
The Complete Master Prompt — Ready to Use
This is the consolidated, production-ready version of everything above. Upload your photo, paste this, and go.
Using the uploaded photo as reference, create a Japanese shojo manga-style illustration with these exact characteristics: LINE ART: Clean ink outlines with visible weight variation — bold on character edges, finer on interior details and background. Confident, intentional strokes. Small ink splatter dots on the ground. COLORING: Selective fill only — warm natural skin tones on both people, clothing colors matching the photo, soft green grass beneath them. Everything else (trees, sky, background figures) remains pencil sketch / greyscale. No background color fill. CHARACTERS: Two people sitting side by side, laughing and looking at each other. Match the hair color, skin tone, clothing style, and general face shape from the reference. Expressive, warm smiles — natural not cartoonish. The woman has long hair, the man has short darker hair and a beard. DECORATIVE OVERLAYS: Float these hand-drawn doodles around the couple — small pink hearts (3–4), yellow and blue five-pointed stars (4–5), eighth note and quarter note music symbols (3–4). Slightly irregular, like a real person drew them with a marker. Scattered naturally, not symmetrical. BACKGROUND: Tall tree trunks sketched loosely, pine branch sprigs at the top corners, 2–3 silhouetted figures walking in the far background (sketch only, no color). Airy composition — background never competes with the foreground subjects. MOOD: Warm afternoon, park setting, slice-of-life manga — like a scene from a romance webtoon. Playful, light, emotionally warm. OUTPUT: Square format (1:1), portrait orientation, no text or watermarks, no UI elements.
Style Variations — Beyond the Classic Look
The image above uses one specific style. Once you've nailed the base, these variations let you create completely different moods from the same photo.
Prompt Tips That Actually Make a Difference
Don't say "anime style" — that's too broad and gives you randomness. Say "shojo manga," "Korean webtoon," "slice-of-life manga illustration," or "Japanese comic art." The more specific the genre, the more consistent the output.
Selective coloring is what makes this image stand out. You have to explicitly say "background is pencil sketch only, no color fill on trees or sky." If you don't say this, will color everything.
Don't say "add some cute stuff around them." Say "floating hand-drawn pink hearts (3–4), yellow and blue five-pointed stars (4–5), music notes." Count matters. Color matters. The more specific, the more control you have.
"Like a scene from a romance webtoon" or "in the style of Takehiko Inoue's line work" gives a cultural context anchor. It narrows the aesthetic universe dramatically. Use this even if the AI can't perfectly replicate named artists.
Don't open a new chat to regenerate. Keep the same conversation thread alive — ChatGPT retains context of your previous generations and your uploaded photo. Type "keep the same style but..." and it continues from where it left off. Starting fresh loses all that context.
ChatGPT can't guarantee exact facial likeness — it's not trained to clone specific people. What it captures well: skin tone, hair color, hair length, body proportions, and general energy. If likeness is critical, try Midjourney v7 or Adobe Firefly with reference image weighting — they have more advanced reference controls.
Use Cases — What to Do With the Final Image
- Custom Gift — print on canvas, a mug, phone case, or art print. Print-on-demand sites like Printify or Canva Print accept high-res PNGs. A personalized manga portrait makes a genuinely memorable gift.
- Couple Profile Picture — matching social media profile photos that actually look artistic instead of just filtered selfies.
- Anniversary Card — export to Canva, add a date and message, print or send digitally. Ten minutes total.
- Wedding Decor — the illustration style works beautifully for save-the-date cards, table cards, or ceremony programs.
- Social Content — post the before/after (real photo + illustration) on Instagram or TikTok. This kind of AI transformation content consistently performs well.
- Webtoon / Story Illustration — if you write stories, use this as character art for your couple without needing to commission an artist.
The most expensive part of a handmade illustration used to be the artist's time. Now the time cost is 20 minutes and the right question.
Quick Recap — Everything You Need
- ChatGPT Plus — 4 handles reference-based illustration
- Upload a clear, well-lit reference photo of both people
- Specify line art style, selective color, and doodle elements precisely
- Name the genre: "shojo manga" or "slice-of-life webtoon"
- Iterate in the same chat thread — don't restart
- Upscale with Topaz or Firefly for print-ready resolution
- Finish in Canva for text, branding, or card layouts