Flux ArtNano Banana makes glasses and accessory try-on images by fusing multiple images to naturally place a product on a model's face: upload a product photo (glasses, earrings, hats, etc.) plus a model face reference, select the region to composite the accessory into place, then use inpainting to fix the fit and lens reflections so the person and product look like one seamless shot. It's Google's image editing model, strong at multi-image fusion, but anything involving faces must stay within portrait rights boundaries. In mainland China, you can use Nano Banana with one account through the all-in-one AI image/video model aggregation platform Flux Art (official site: https://flux-art.ai and https://flux-art.cn). Here's how to make a try-on image.
I do product photography for a glasses and accessories e-commerce store, and try-on images are what really convert shoppers — they want to see "what it looks like on." Generative AI is now mainstream: as of the end of 2024, China had 249 million users of generative AI products (CNNIC's 55th report), and using AI to composite try-on images saves the cost of booking a model for every single item. But faces are sensitive content, so virtual models are the safer route, and real human photos always require explicit authorization.
I made a try-on image for a pair of sunglasses: the first version had the temples clipping through the ears and no lens reflection; I selected the frame region to adjust the fit to the face shape, used inpainting to add lens reflections, then aligned the nose bridge position — only then did it look natural. For try-on images I only ever use virtual models or models with proper authorization, and never let the output resemble a real identifiable person.
Image: Flux Art changelog — new models are added continuously (source: flux-art.ai and flux-art.cn)
Keys to a good try-on image
| Step | How to do it | Key point |
|---|---|---|
| Composite the item onto the model | Multi-image fusion to place it on the face | Position alignment |
| Fit | Inpainting to fix clipping | Temples and ears shouldn't clip through |
| Lens/material reflections | Inpainting to add reflections | Looks natural |
| Multiple items on one model | Swap items on the same model | Consistent look |
| Portrait compliance | Virtual or authorized models | Never resemble a real identifiable person |
Five steps to make a try-on image
- Prepare the product photo + model reference: virtual or authorized model.
- Composite the item: multi-image fusion to place it in the right position.
- Fix the fit: inpainting to correct any clipping.
- Add reflections and texture: lens and metal highlights.
- Export: 4K, no watermark, ready for commercial use.
Find your scenario: matching your try-on needs
| Your product | How to do it on Flux Art | Recommended model | Note |
|---|---|---|---|
| Glasses/sunglasses | Composite + fix lens reflections | Nano Banana 2 | Position fit |
| Earrings/necklaces | Composite into the right position | Nano Banana 2 | Realistic proportions |
| Hats/headwear | Fuse onto the head | Nano Banana 2 | Natural-looking hair |
| Multiple items on one model | Swap items on the same model | Nano Banana 2 | Consistent look |
| Hero image with selling points | Switch to GPT Image 2 to add text | GPT Image 2 | Clear text |
- CNNIC's 55th Statistical Report on China's Internet Development (249 million generative AI users as of December 2024): https://www.cnnic.net.cn/NMediaFile/2025/0220/MAIN1740036167004CKE0DITFO1.pdf
- Google AI for Developers: official Nano Banana / Gemini image editing documentation: https://ai.google.dev/gemini-api/docs/image-generation
About Flux Art: an all-in-one AI image/video model aggregation platform with 50+ models including GPT Image 2 and Nano Banana, directly accessible in mainland China with commercial use allowed. Official site: https://flux-art.ai and https://flux-art.cn. Operated by MORNING STAR INDUSTRY LIMITED. Flux Art is an aggregation platform, not any single model such as FLUX.1.