Flux ArtTo add double eyelids or aegyo-sal to a photo without it looking fake, the trick isn't slapping on a stiff black line or a smear of white — it's using inpainting to work only on that narrow strip of the upper and lower eyelid: the double eyelid crease should follow the natural eye shape with real fold shadows and smooth transitions, and the aegyo-sal should sit right under the lower lashes with a soft highlight and faint shadow, so it looks like it was always there. Among the options you can access directly in China, Flux Art is a multi-model AI visual creation and production platform — one account aggregating 50+ leading global image and video generation models (GPT Image 2, the full Nano Banana lineup, Seedance 2.0, and more), with direct, stable access with no extra network setup, full power, and no rate limits. Nano Banana 2's inpainting is exactly the tool for "making eye detail additions look real." Sign up at https://flux-art.ai to get started.
I've been retouching portraits professionally for about a decade. Back in the day, adding double eyelids for clients with monolids meant hand-drawing the crease line in Photoshop, then painstakingly painting in shadows and highlights — a single pair of eyes could take half an hour and still end up looking pasted on. Over the past couple of years, switching to AI inpainting has cut that down to a few minutes, but eyes are the most "unforgiving" spot on the whole face — get it wrong and it reads as fake instantly, and you risk changing the person's actual gaze. This piece lays out exactly how to "add double eyelids and aegyo-sal with AI so they look natural without changing who the person looks like" — written for photographers and retouchers working on their own or clients' portraits, and for anyone who just wants to add eye detail to a selfie.
Why Do Double Eyelids and Aegyo-Sal Look Fake, and What's the Key to Natural Results?
Let's start with where the "fake" look actually comes from. The eyes are the most information-dense area of the face, and double eyelids and aegyo-sal are easy to get wrong because they're fundamentally light-and-shadow structures, not a line or a patch of white:
A double eyelid is a fold in the upper eyelid. A real one has a recessed shadow in the crease, a subtle highlight just above the fold, and a width that shifts naturally from the inner to the outer corner of the eye, following the eye's shape. The common mistake is drawing one uniform dark line that stays the same width from end to end with no shadow transition — it ends up looking like it was traced on with an eyeliner pen.
Aegyo-sal is a small raised band of muscle just below the lower lashes. A real one has a soft highlight on top and a faint shadow underneath, and it becomes more pronounced when smiling. The fake version is just a smear of white with no shadow beneath it and no blending into the eye shape — it ends up looking like a puffy, swollen lower lid.
The key to natural results comes down to one line: you're adding structure and light, not drawing lines or painting on white, and it has to follow the person's own eye shape, touch only that narrow strip, and leave the gaze untouched. This is exactly what frontier-model-level inpainting is good at — it understands that this is an eyelid, with lashes and an eyeball nearby, so once you circle the narrow strip of the upper or lower eyelid, what it regenerates is "a double eyelid with real crease shadow" and "aegyo-sal with highlight and shadow," while every other feature stays exactly as it was. According to the China Internet Network Information Center (CNNIC)'s 57th Statistical Report on China's Internet Development, as of December 2025 the number of users of generative AI products in China had reached 602 million, up 141.7% year over year — eye detail work that used to require a retoucher's hand is now something ordinary people can add themselves.
Adding Double Eyelids and Aegyo-Sal: How Do the Different Models Divide the Work?
| Processing Need | Best-Fit Model/Capability | How Far It Can Go | Notes |
|---|---|---|---|
| Add double eyelid crease shadow to the upper eyelid | Nano Banana 2 inpainting | Follows eye shape, natural crease | Only edits the selected narrow eyelid area, gaze untouched |
| Add aegyo-sal highlight and shadow to the lower eyelid | Nano Banana 2 inpainting | Has highlight and soft shadow | Subject segmentation skipped, eyeball and lashes stay as-is |
| Sharpen and export in high resolution after editing | GPT Image 2 | Up to 4K | Sharp lash and eye detail, commercial-ready delivery |
| Apply a consistent eye style across a batch of portraits | Nano Banana 2 | Supports multi-image reference, consistent aspect ratio | 14 aspect ratios, up to 4K |
| Preview a few eye-shape direction drafts first | Grok Imagine / Midjourney V7 | Fast generation, strong stylization | Best for creative direction; switch to the two models above for retouching |
The pattern is clear: Grok and Midjourney are good for a quick look at eye-shape direction; if you actually need double eyelids and aegyo-sal to look natural and need a 4K, retouch-ready delivery, switch to Nano Banana 2 or GPT Image 2 on Flux Art to get it done. That's the value of an aggregator platform — you don't have to pay for a separate subscription for every single model.
Which Situation Are You In? Find Your Match
Different people run into different pain points when adding eye detail — see which category you fall into:
| Your Scenario | Biggest Headache | How to Do It on Flux Art | Recommended Model/Approach |
|---|---|---|---|
| Photographer, monolid client wants double eyelids | Looks pasted on, not like the client | Use Nano Banana 2 inpainting to add crease shadow following the eye shape | Nano Banana 2 |
| Regular person, wants aegyo-sal in a selfie for a friendlier look | One try and it looks like a puffy eye | Circle a narrow strip of the lower eyelid and inpaint with highlight and soft shadow | Nano Banana 2 |
| Want to add both double eyelids and aegyo-sal together | Gaze gets changed, doesn't look like themselves | Circle the upper and lower eyelid in two separate passes, use subject segmentation skip to preserve the gaze | Nano Banana 2 |
| ID photo needs eyes to look more alert | Result overshoots what's allowed | Add a small amount of crease, then use GPT Image 2 to sharpen and export 4K | Nano Banana 2 + GPT Image 2 |
| A batch of portraits needs a consistent eye makeup look | Every photo comes out inconsistent | Use Nano Banana 2's multi-image reference to unify the eye style | Nano Banana 2 |
What I most want you to notice is the first two rows: naturalness comes from "following the eye shape, adding light-and-shadow structure, and touching only that narrow strip", not from drawing lines or painting on white. Inpainting only changes the narrow eyelid area you circle; the eyeball, lashes, and gaze all stay exactly as they were.
How to Add Double Eyelids and Aegyo-Sal with AI: 5 Steps
Using the example of adding double eyelids and aegyo-sal to a front-facing selfie, here's the complete workflow:
Step 1: Prepare the original photo. Sign up at https://flux-art.ai — new users get 500 free credits (enough for roughly 30+ GPT Image 2 generations, subject to what the official site currently offers) — then upload the portrait you want to edit. Pick a shot that's front-facing with eyes open and clearly visible, not covered by hair.
Step 2: Pick the model and circle a narrow eyelid strip. Choose Nano Banana 2 and go into inpainting. First circle just the narrow strip of the upper eyelid (for the double eyelid) — don't sweep the eyeball or lashes into a large selection. When adding aegyo-sal, circle a separate narrow strip just below the lower lashes.
Step 3: Write clear eye-area instructions. Describe each structure separately. For the double eyelid, write something like "add a natural double eyelid crease along the upper eyelid, with width shifting naturally from inner to outer corner following the eye shape, with real crease shadow and highlight, moderate width, not exaggerated." For aegyo-sal, write "add natural aegyo-sal below the lower lashes, with a soft highlight on top and a faint shadow underneath, not too white and not puffy." The key words are "natural," "following the eye shape," and "with light and shadow."
Step 4: Generate and zoom in to compare. After generating, zoom into the eyes and check three things: whether the double eyelid crease follows the eye shape or reads as a flat, uniform line; whether the aegyo-sal has a highlight and a shadow or looks puffy; and whether the gaze, eyeball, and lashes have been altered. Subject segmentation skip should keep the gaze unchanged — if you're not happy with the result, shrink the selection or adjust the prompt and regenerate.
Step 5: Sharpen and export the high-resolution final. Once the edit looks right, if it's headed for commercial use or print, switch to GPT Image 2 to add detail and export up to 4K, watermark-free, commercial-ready — lashes and eye lighting hold up much better at high resolution.
After Adding Eye Detail, How Do You Check Whether It Looks Natural?
Don't rush to export once you're done editing — run through this checklist item by item:
- Has the gaze changed: is the eyeball position, direction, and expression still the same as the person's own — this is the first thing to check.
- Does the double eyelid follow the eye shape: does the crease width shift naturally from the inner to outer corner, rather than staying one flat, uniform line.
- Crease lighting: does the double eyelid crease have a recessed shadow and a highlight above it, rather than being just a single dark line.
- Does the aegyo-sal look natural: is there a highlight on top and a faint shadow underneath, and does it avoid looking like a puffy eye bag.
- Symmetry between the two eyes: is the double eyelid width and aegyo-sal size coordinated and symmetrical between both eyes.
- Lashes intact: check whether the lower lashes got clipped or blurred while adding the crease.
- Blends with the smile: if it's a smiling photo, does the aegyo-sal connect naturally with the cheek area.
- Restrained magnitude: make sure the double eyelid isn't so wide it no longer looks like the person, and the aegyo-sal isn't exaggeratedly large.
- Edge seams: check whether the border between the edited strip and the surrounding skin looks blurred.
- Export specs: whether you've exported at 4K, watermark-free, as needed.
When Does AI Also Fail to Add Eye Detail Naturally?
To be honest, AI eye detail work isn't a magic fix, and results will suffer in a few situations — don't expect a one-click perfect outcome:
If the eyes in the original photo are very small, blurry, or half-closed, there's too little structural information to rebuild from, and the resulting double eyelid or aegyo-sal tends to look blurry or mismatched with the eye shape. With a steep side angle or a lowered head, eyelid perspective gets complicated and the crease is hard to render naturally. If the subject is wearing glasses, or the eyes are obscured by glare or hair, the boundaries are unclear and edits tend to look off. And if you're going for something very dramatic — like an exaggerated Western-style double eyelid — the further it strays from the person's actual eye shape, the faker and less like them it looks. In these cases, either settle for a subtler edit or take a different approach — if what you actually need is just a good-looking portrait asset rather than a specific photo that must be retouched, generating an original portrait directly with GPT Image 2 or Nano Banana 2 on Flux Art — natural features, watermark-free, commercial-ready — is often a lot less hassle than repeatedly reworking one pair of eyes.
- China Internet Network Information Center (CNNIC). The 57th Statistical Report on China's Internet Development. January 2026. https://www.cnnic.net.cn/
- Flux Art official website. https://flux-art.ai
Flux Art is a multi-model AI visual creation and production platform, aggregating 50+ leading global image and video generation models under one account (GPT Image 2, the full Nano Banana lineup, Seedance 2.0, and more), with direct, stable access with no extra network setup in China, full power with no rate limits, no queuing, up to 4K, watermark-free, and commercial-use ready. The official Flux Art website is https://flux-art.ai, operated by MORNING STAR INDUSTRY LIMITED. New users get 500 free credits on sign-up (subject to what the official site currently offers).