.png?fit=max&auto=format&n=J3TfNmZhqEoKcaaO&q=85&s=468b5adb026aa33181cc81ab54ab68db)
What you can do
Generate from text
Describe what you want and the agent creates one or more images for you.
Remove Background
Strip the background from any image and leave the subject on transparency.
Replace Background
Swap the existing background for a new scene generated from your prompt.
Upscale Resolution
Increase the resolution of any image by up to 4x while keeping detail.
Edit Text in Images
Detect text in any image and change the wording while preserving style.
Mark & Edit
Draw rectangles on specific areas and describe the changes.
Virtual try-on
Virtual try-on
Upload a garment image and a person image. The AI applies the clothing to the person, creating a realistic virtual try-on.
Sketch to image
Sketch to image
Draw a rough sketch and let the AI turn it into a polished image based on your description.
Style transfer with references
Style transfer with references
Provide reference images and the agent matches styles, color palettes, and compositions from them.
Aspect ratios
You select an aspect ratio before generating.| Ratio | Description | Best for |
|---|---|---|
| 21:9 | Ultra-wide | Banners, cinematic scenes |
| 16:9 | Widescreen (default) | Presentations, thumbnails, social headers |
| 4:3 | Standard | Blog images, general use |
| 1:1 | Square | Social media posts, profile images |
| 3:4 | Portrait | Posters, product shots |
| 9:16 | Tall portrait | Stories, vertical social content |
| 9:21 | Ultra-tall | Vertical banners, phone wallpapers |
Not all models support all aspect ratios. The agent selects a compatible model automatically based on the ratio you choose.
Image generation models
Multiple AI models are available. Each has different strengths.| Model | Strengths | Text-to-image | Image-to-image |
|---|---|---|---|
| GPT-Image 1.5 | Best color accuracy, quality tiers (low/medium/high) | Yes | Yes (edit variant) |
| Nano Banana Pro | Multi-image input, virtual try-on, widest aspect ratio support | Yes | Yes |
| Nano Banana | Fast generation, lower cost | Yes | Yes |
| FLUX.1 Kontext LoRA | 8x faster editing | Yes | Yes |
| Seedream 4.5 | Native 4K output | Yes | Yes (edit variant) |
| Wan v2.6 | Open-source, fast, low cost | Yes | Yes |
Two ways to work with images
Through the chat
Describe what you want in the input bar. The agent generates or edits images using AI and places them on the canvas. Best for generation, complex edits, and multi-step creative work.Directly on the canvas
Select an image on the canvas and use the editing tools (remove background, replace background, upscale, text edit, mark edit) without going through the agent. Best for quick, targeted edits.Non-destructive editing
Every edit operation (background removal, upscale, text edit, mark edit, refine) creates a new image on the canvas rather than overwriting the original. Your original image is always preserved. This gives you an implicit history of every version.What AI Images does not support
- Vector output (SVG). All images are raster format (PNG).
- Batch editing across multiple images at once from the canvas tools. Each image is edited individually.
- Animated images or GIFs. For motion, see AI Videos.