Three examples
A new light
Before After The ask
Make it evening, with the street lamps on and the sky a deep blue.
Make it evening, with the street lamps on and the sky a deep blue.
A new setting
Before After The ask
Put the bottle on wet black stone instead of the kitchen table.
Put the bottle on wet black stone instead of the kitchen table.
A new size
Before After The ask
Make a vertical version for Reels.
Make a vertical version for Reels.
Three steps, in one chat.
Add your picture
Press + in the chat, upload it and pick Reference. Or press Edit on a picture on your board: it goes to the Director as a reference.
Say what to change
The Director writes the Prompt with your picture as its reference. A crop or a new size it makes at once, for free.
Generate
The card shows the model and the price on Generate. No render is charged until you press it.
28 models can change your picture.
Two of them are here, side by side. They take your picture as a reference, and make a new one from it and your words.
| Model | Takes your pictures | Largest | Starts at |
|---|---|---|---|
| Nano Banana 2.1 | Up to 10 | 4K | 3 creditsone image at 1K |
| Nano Banana Pro | Up to 8 | 4K | 14 creditsone image at 2K |
FLUX.2 Pro and GPT Image 2.5 Flare make pictures from words alone, so they can't take yours. Each starting price is one new picture, at the size the model opens on.
The practical details
AI Image Editor, in practice
- What it takes
- Your picture as a reference: a JPG, PNG or WebP, at least 320 pixels on each side. Then your words: what should change.
- What it makes
- A new picture, made by the model from yours and your words. It's drawn again, not retouched, so details you didn't mention can come out different.
- Exact edits
- Crop to a shape or an area, change the type to JPG, PNG or WebP, or cut a still from a video. Each runs at once, generates nothing and costs nothing, and the file is kept in your uploads.
- A platform's size
- Vertical 9:16 (TikTok, Reels, Stories, Shorts), 1080 × 1920. Portrait 4:5 (a feed post), 1080 × 1350. Square 1:1 (a feed post), 1080 × 1080. Widescreen 16:9 (YouTube, a thumbnail), 1920 × 1080. Each fills the frame, cutting its edges, or keeps all of it with bars.
- What it costs
- A change by words is priced as one picture on its model, at its size. Qwen 3 Image to image and Qwen 3 Pro image to image add a charge for each picture of yours; on the others, your picture adds nothing. An exact edit is free.