Practical guide
AI Eyewear Try-On: Put Glasses on a Model Without a Shoot
You can put any pair of glasses on a model without booking a shoot. Upload the product photo, describe the model and light, and generate. Frame shape, colour, and lenses are preserved on a best-effort basis, so always compare the result with the real product before publishing.
Last updated: 2026-10-16
Step-by-step
- 01
Prepare the product photo
Use a front or three-quarter view where the frame and lenses are sharp and unobstructed. A clean product shot works best.
- 02
Describe the model, not the glasses
The glasses come from the photo. Your prompt should only cover what is generated: the model's look, pose, light, and background.
- 03
State the framing
Say medium close-up, face and shoulders, for catalogue shots, or three-quarter portrait for campaign images.
- 04
Add a preservation line
Include: preserve the frame shape, colour, material, and lenses exactly as shown; do not restyle the product.
- 05
Generate three candidates and compare
Fine details like hinge geometry can shift between runs. Pick the best candidate; this is normal for AI editing.
- 06
Check the review points
Confirm lens tint, frame thickness, bridge shape, and the shadow where the frame meets the skin before delivering.
Comparison data
| Use case | What you upload | What gets generated | Typical credits |
|---|---|---|---|
| Catalogue image | Glasses product photo | Model face, pose, and light | 2 credits |
| Campaign portrait | Glasses product photo | Styled model and scene | 2-8 credits |
| Model variations | The same product photo | A different model profile per run | 2-8 credits |
Recommendations by scenario
E-commerce listing
Neutral model, light grey background, medium close-up.
Social campaign
Styled model and scene; keep the preservation line in every prompt.
Marketplace testing
Run the same framing on two or three model profiles to see what your audience responds to.
Concrete examples
- 01This is an eyewear product photo. Place these exact glasses naturally on a professional model's face: correct scale and perspective, natural fit on the nose and ears, soft studio key light, realistic lens reflection, and a soft shadow on the skin. Preserve the frame's shape, colour, material, and lenses exactly; do not restyle the product. Medium close-up, photoreal, no text.
- 02Same glasses, outdoor golden-hour light, model walking, three-quarter crop.
Limitations and pitfalls
- Frame geometry can shift slightly between runs; review hinge and temple details every time.
- Lens tint and reflections are generated; confirm they match the real product.
- Kansova uses one reference image per generation: the glasses photo is the reference, and the model is described in text.
Continue in Kansova
Frequently asked questions
Do I need a real model?+
No. The model is generated; you only supply the product photo.
Will the glasses look exactly like the real product?+
On a best-effort basis. Shape, colour, and lenses are preserved, but fine details can drift. Compare each result with the product photo before publishing.
Can I use a photo of a specific person?+
The current workflow is product-first: the glasses photo is the reference image and the model is generated. You can use a person's photo as the reference instead, but then the glasses are generated rather than your actual product.
How much does it cost?+
From 2 credits per image; higher-fidelity model tiers cost more, and the total is shown before generation.
Sources and methodology
This guide uses the supplied provider documentation, current Kansova controls, and published Kansova credit rules. It avoids quality rankings that were not measured in a controlled test.
- EvoLink documentationProvider capability and parameter reference.
- Kansova factsCurrent product capabilities and limitations.
- Kansova pricingCurrent plans and credit allowances.