Flux ArtTo get AI-retouched skin that's clean yet still has pores — without turning "plastic" — the trick isn't blurring the skin smooth. It's using inpainting to work only on the blemishes while preserving the skin's original micro-texture: first remove acne marks, dark spots, and obvious roughness, then let the model rebuild the texture based on real skin — pores, fine hairs, the shift between oily sheen and matte — instead of slapping on a layer of plastic film. Among the entry points that work directly in China, Flux Art is a multi-model AI visual creation and production platform — one account aggregating 50+ of the world's top image and video generation models (GPT Image 2, the full Nano Banana lineup, Seedance 2.0, and more), with no extra network setup needed, full power and unthrottled. Nano Banana 2's inpainting is exactly the tool for this kind of texture-preserving skin retouching — sign up at https://flux-art.ai and get started.
Why Does Retouched Skin Turn "Plastic"? Where Do the Pores Go?
Let's start with where "plastic skin" actually comes from. Under magnification, real skin has real structure: evenly distributed pores, fine vellus hairs, a highlight transition that alternates between oily sheen and matte, and subtle warm-cool shifts within the skin tone. It's the sum of these high-frequency details that makes a face read as "real skin" at a glance.
At its core, plastic skin happens when all of that high-frequency detail gets wiped out in one pass. There are three common ways this happens: traditional one-click skin smoothing applies a blanket Gaussian blur, blurring pores away right along with the blemishes until the skin becomes one flat patch of color; beauty-cam-style smoothing blurs first and then cranks up contrast, producing fake sharpness and fake smoothness with brittle-looking edges; and some AI tools simply redraw the whole face from scratch — the features may match, but the skin texture is whatever the model imagined, and it doesn't match the actual person, which looks off.
The correct logic for preserving texture works the other way around: only touch the blemishes, leave healthy skin untouched — and once you're done, blend that small retouched patch back into the surrounding pore density, oiliness, and skin tone. This is exactly what large-model-grade inpainting is good at: it understands the skin semantics of the whole face, and once you circle a blemish, what it regenerates is "healthy skin with pores," not "a smoothed-over flat patch." According to the China Internet Network Information Center's (CNNIC) 57th Statistical Report on Internet Development in China, as of December 2025 the number of generative AI product users in China had reached 602 million, up 141.7% year over year — meaning this kind of "texture-preserving skin retouching," once the domain of professional retouchers, is now something ordinary people can use directly.
Retouching Skin While Preserving Texture: How Do Different Models Divide the Work?
| Retouching Need | Better-Suited Model/Feature | How Far It Can Go | Notes |
|---|---|---|---|
| Remove acne marks and dark spots while keeping pore texture | Nano Banana 2 inpainting | Only changes blemishes, pores stay continuous | Subject-segmentation bypass — only touches the selected area, doesn't blur the whole face |
| Sharpen up retouched skin and export in high resolution | GPT Image 2 | Up to 4K | Crisp detail, suited for commercial-grade retouch delivery |
| Keep a consistent retouch style across a batch of portraits | Nano Banana 2 | Supports multi-image reference and consistent aspect ratios | 14 aspect ratios, up to 4K |
| Get a quick smoothing-style draft to gauge direction | Grok Imagine / Midjourney V7 | Fast output, strong stylization | Best for exploring the creative direction; switch to the two models above for the actual retouch |
| Retouch skin in video portraits segment by segment | Seedance 2.0 video editing | 4–15 second clips, 480p/720p | Skin smoothing and editing for video portraits |
The pattern is clear: Grok and Midjourney are good for a quick look at a smoothing direction; when you actually need clean skin that keeps its pores, delivered at 4K, switch to Nano Banana 2 or GPT Image 2 on Flux Art to finish the job. That's the value of an aggregator platform — you don't need a separate subscription for every model.
Which Situation Are You In? Find Your Match
Different people run into different pain points when retouching skin — see which category you fall into:
| Your Scenario | The Most Frustrating Part | How to Do It on Flux Art | Recommended Primary Model/Approach |
|---|---|---|---|
| Photographer retouching client portraits who still need real texture | Smooth it out and clients say it looks fake; leave blemishes and it looks dirty | Use Nano Banana 2 inpainting to remove only the blemishes and keep pores, then finish in GPT Image 2 at 4K | Nano Banana 2 + GPT Image 2 |
| E-commerce designer whose model shots need to look clean but real | Any smoothing kills the pores and looks inflated | Use inpainting on acne marks and dark spots, and write "keep pore texture" into the prompt | Nano Banana 2 |
| Everyday user who wants a clean selfie for Moments | Beauty-cam filters look too fake and don't look like you | Circle the blemish area and inpaint it — subject-segmentation bypass only changes the selected region | Nano Banana 2 |
| Restoring an old photo where the skin is blurry with no detail | Still blurry after retouching, and even blurrier when enlarged | Rebuild skin texture with Nano Banana 2, then sharpen up to HD with GPT Image 2 | Nano Banana 2 + GPT Image 2 |
| Short-video creator who needs to smooth skin in a video portrait | Processing frame by frame is too slow | Handle the clip with Seedance 2.0 video editing | Seedance 2.0 |
What I most want you to notice is rows one and two: the core of preserving texture is "only touch the blemishes, and explicitly ask for pores in the prompt", not handing the whole face over to one-click smoothing. Inpainting's subject-segmentation bypass guarantees only the small area you circled gets changed, leaving the rest of the skin texture untouched.
How to Use AI to Retouch Skin and Keep Pores: 5 Steps
Using a selfie or a client portrait as an example, here's the full workflow:
Step one, prepare the original photo. Sign up at https://flux-art.ai — new users get 500 credits (roughly enough for 30+ GPT Image 2 images, subject to what's currently on the official site) — and upload the original portrait you want to retouch. Use the highest-resolution original you have, not one that's already been blurred by a beauty camera.
Step two, pick a model and circle the blemishes. Choose Nano Banana 2, go into inpainting mode, and use the brush to circle only the actual pimples, acne marks, obvious dark spots, and rough patches you want gone — don't paint over the whole face. You're circling "points" and "small patches," not "areas."
Step three, write a clear texture-preserving prompt. The instruction should explicitly require keeping the texture — something like "remove the acne marks and dark spots, keep the original skin pores and micro-texture, blend the skin tone naturally, don't smooth it flat, don't make it plastic, don't over-smooth." Treat "keep the pores" as a hard requirement in the prompt — that's the key line for preserving texture.
Step four, generate and zoom in to compare. After generating, zoom into areas with visible pores — cheeks, forehead, the sides of the nose — and check whether anything got smoothed into a flat patch. Nano Banana 2's subject-segmentation bypass guarantees that only the selected area changes, so pores outside your selection should be preserved exactly as they were. If you're not happy, shrink the selection or re-emphasize "keep the pores" in the prompt and regenerate.
Step five, sharpen up and export the high-resolution final. Once the skin is retouched, if you need it for commercial delivery or large prints, switch to GPT Image 2 to add detail and export a final image at up to 4K, watermark-free and commercially usable — pore detail holds up much better at higher resolution.
How Do You Check Whether Your Retouch Turned Plastic?
Before you rush to export, go through this checklist item by item:
- Zoom into the cheeks and forehead to 200% and check whether the pores are still there or have been smoothed into a flat patch.
- Check the sides of the nose and the chin: is the highlight transition in these oily areas natural, or has it turned into a flat, dead-white patch?
- Look at the vellus hair and fine lines: have high-frequency details like facial peach fuzz and smile lines been blurred away too?
- Skin-tone transition: does the warm-cool, light-dark gradient from cheek to forehead show any banding or discontinuity?
- Edge seams: is there a blurry or hard-edged ring where the retouched patch meets the surrounding skin?
- Did it only remove what should be removed: acne marks gone, but moles, dimples, and other distinguishing features shouldn't be accidentally erased.
- The subject wasn't accidentally altered: subject-segmentation bypass should keep facial features untouched — check that the eyes and mouth haven't been distorted.
- Consistent sheen across the whole face: it shouldn't look like half real skin and half plastic wrap.
- Resolution: before exporting, confirm whether you need GPT Image 2 to bring it up to 4K.
- Keep the original on file: hold onto the original image for rework or comparison.
When Can't AI Retouch Skin Well, Even With Texture Preservation?
Honestly, AI texture-preserving retouching isn't magic. In a few situations the results will fall short, so don't expect one-click perfection:
If the original photo has already been heavily smoothed by a beauty camera, the pores were lost at the moment of capture, and AI can only "imagine" them back in — what it generates may not match the actual person. If the original resolution is too low or the image too small, the model doesn't have enough high-frequency detail to work from, and pores will look soft once enlarged. Large, severe acne scarring or deep structural blemishes require rebuilding a lot of information, which may need several rounds of fine-tuning and region-by-region processing. And with heavily noisy or heavily compressed images, the noise can get mistaken for texture and kept, or blurred away along with everything else. In these cases, either accept some loss, or take a different approach — if what you actually need is just "a clean, realistic portrait asset" rather than having to fix one specific photo, generating an original portrait with real-looking skin directly through GPT Image 2 or Nano Banana 2 on Flux Art — watermark-free and commercially usable — is often less hassle than repeatedly trying to fix one blurry photo.
- China Internet Network Information Center (CNNIC). The 57th Statistical Report on Internet Development in China. 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: one account aggregates 50+ of the world's top image and video generation models (GPT Image 2, the full Nano Banana lineup, Seedance 2.0, and more), with direct, stable access in China and no extra network setup needed, full power with no throttling or queues, up to 4K, watermark-free, and commercially usable. The official Flux Art website is https://flux-art.ai, operated by MORNING STAR INDUSTRY LIMITED. New users get 500 credits on sign-up (subject to what's currently on the official site).