Canvas guide
The Canvas is where you write your own prompt, render, and then edit the results. It has two parts: the prompt bar at the bottom, and the image viewer that opens when you click an image.
The prompt bar
- Add images: drop them on the page, paste one with Ctrl+V, or click Images. Up to 6 images per run.
- Describe the result in your own words, in any language. Type @ to cite one of your images by its alias (for example
@facade). - Pick the settings in the row under the field: model, format, resolution, number of images.
- Press Ctrl+Enter (⌘+Enter on Mac) or click Generate. The cost shows on the button before you click.
Results land on the wall of your project (or in the scene shown). Switch to Feed in the toolbar to see your runs stacked, newest at the bottom.
The Image · Video switch at the left of the settings row turns the bar into an Animate bar. See Video.
Templates in the bar
The Templates chip runs a template right from the bar, with your images and its options. See Templates.
Models
| Model | Plans | Good for |
|---|---|---|
| Nano Banana Pro (Google) | All plans | Maximum detail and fidelity to your source image, several references |
| Nano Banana Flash (Google) | All plans | Faster and cheaper, with extra-wide formats |
| GPT Image 2.5 (OpenAI) | Paid plans | Text and image to image. Two variants: Flare (faster, for everyday renders) and Sunburst (the most precise edits of your source) |
Pick the model in the Model chip. Switch between Pro and Flash, or between Flare and Sunburst, in Advanced settings (the gear chip). On GPT Image 2.5, Advanced settings also has a Quality setting (Low to Max), which changes the cost.
Images made with a model ARKIA no longer offers keep their original label in your gallery.
Formats and resolutions
| Nano Banana Pro | Nano Banana Flash | GPT Image 2.5 | |
|---|---|---|---|
| Formats | Auto, 1:1, 16:9, 9:16, 4:3, 3:4, 2:3, 3:2, 4:5, 5:4, 21:9 | The same, plus 1:4, 4:1, 1:8, 8:1 | Auto, 1:1, 2:3, 3:2, 3:4, 4:3, 4:5, 5:4, 9:16, 16:9, 21:9 |
| Resolutions | 1K, 2K, 4K | 1K, 2K, 4K | Auto, 1K, 2K, 4K |
- Auto format follows your source image; it shows once an image is in the bar.
- Auto resolution (GPT Image 2.5) picks the size closest to your source.
- Images per batch: 1 to 4.
Advanced settings
The gear chip at the end of the settings row:
| Setting | What it does |
|---|---|
| Variant | Pro or Flash (Nano Banana), Flare or Sunburst (GPT Image 2.5) |
| Quality | GPT Image 2.5 only, Low to Max |
| Seed | Nano Banana only. -1 draws a random seed, saved with each image. Same seed, close result. GPT Image 2.5 has no seed. |
| Grounding | Nano Banana only. Anchors the render on real-world data (Google Search). Adds a few credits per image. |
| All-ref | Nano Banana only. Sends every image as a reference, not only the ones your prompt cites. |
| Chain | Nano Banana only. Keeps the prompt history across successive iterations. |
Enhance prompt
The wand in the bottom-right corner of the field rewrites your prompt: it adds materials, light and camera in the ARKIA archviz style. Shortcut: Ctrl+Shift+E.
After an enhance, the corner shows Original | Enhanced: switch between the two versions before you generate. The wand again enhances from your original prompt.
Enhance costs no credits. Each plan has a usage limit; the wand's tooltip says how many enhances are left when you get close to it.
In Video mode, the wand turns your words into a precise camera move. See Video.
The image viewer
Click any image to open it full size. ← and → move between images, ↑ and ↓ browse the edit history of the image, Esc closes.
Edit tools
The toolbar at the top of the viewer:
| Tool | What it does |
|---|---|
| Edit | Describe a change to the whole image, in your own words. |
| Annotate | Draw on the image to show what to change: Pen, Arrow, Box, Circle, Text, plus Comment and Eraser. Add a few words if you like, then send. |
| Comment | Pin numbered comments on the image, each saying what to change in that spot. |
| Select | Mark an area, then describe the change there or pin comments in it. Tools: Select object (click an object, or type a few words to select every match), Brush, Lasso, Eraser. |
| Erase | Brush over something to remove it. Leave the field empty to erase it, or describe what to put there instead. |
| Remove background | Keeps the subject on a transparent background, as a new edit of the image. 1 credit. |
| Resize | Extends the image to another format: Square (1:1), Portrait (3:4), Story (9:16), Landscape (4:3) or Widescreen (16:9). The cost of each format shows in the menu. |
Each edit is saved as a new version of the image. Find it with ↑ / ↓ in the viewer.
The edit bar works like the main prompt bar: its own model and settings, @ to cite up to 4 more images, and the wand to enhance your instruction.
Select object is free for your first 50 selections each month, then 1 credit per selection.
Actions under the image
| Action | What it does |
|---|---|
| Upscale | Enlarges the image: Free 2× on your own graphics card, or Quick in the cloud (faithful, about 20 s, from 2 credits, up to 4K). See Upscale. |
| Compare | A before / after slider with the source image. |
| Copy | Copies the image, to paste it anywhere. |
| Iterate | Reuses this image as the source of your next run (R). |
| Vary | Loads the prompt, settings and sources into the bar with a new seed, to adjust before generating (V). For a template result, it goes back to the template with the same images and options. |
| Nodes | Sends the image to Nodes. |
| Animate | Turns the image into a short video (A). See Video. |
At the top right: Download, Add to favorites (F), Publish to community, and Delete behind the ⋯ menu.
Upscale
Click Upscale under the image and pick a level:
- Free 2×: runs on your own graphics card, in a few seconds. Needs a browser with WebGPU, such as a recent Chrome or Edge, and stops at 8,192 px on a side.
- Quick: a faithful upscale in the cloud, in about 20 seconds.
Quick picks the factor for you: the largest that stays within 4K (about 17 megapixels), up to ×4. Each option shows the output size; Quick shows its cost and asks you to confirm. An image that's already 4K has nothing left to enlarge: both options show Already 4K.
The result is saved as a new edit of the image, named for example Upscale 4× · Quick, and you can keep working while it runs. If your connection drops, the app says Still processing. It will land in your gallery. The upscale goes on, and if it fails, your credits are refunded automatically. Prices are in Credits and Plans.
Keyboard shortcuts
Press ? in the app to see this list for the screen you are on. On a Mac, Ctrl is ⌘.
Prompt bar
| Shortcut | Action |
|---|---|
| Ctrl+Enter | Generate, or animate in Video mode (runs the template in template mode) |
| @ | Cite one of your images |
| M | Video mode: open the Motion catalog |
| Ctrl+V | Paste an image |
| Ctrl+Shift+E | Enhance prompt |
| Esc | Shrink the expanded prompt |
Wall (on the hovered image)
| Shortcut | Action |
|---|---|
| R | Reuse as source |
| V | Vary: same settings, new seed |
Image viewer
| Shortcut | Action |
|---|---|
| ← / → | Previous / next image |
| ↑ / ↓ | Browse the edit history |
| Esc | Close the tool, then the viewer |
| R | Reuse as source |
| V | Vary: same settings, new seed |
| F | Add to / remove from favorites |
| A | Animate: turn the image into a video |
| M | Open the Motion catalog (Animate) |
| Ctrl+Enter | Send the edit, or launch the video |
Inside the Select and Erase tools
| Shortcut | Action |
|---|---|
| S | Select object |
| B | Brush ([ and ] resize it) |
| L | Lasso (Enter closes a shape drawn point by point) |
| E | Eraser |
| C | Comment |
| Ctrl+Z / Ctrl+Shift+Z | Undo / redo |
Costs
The cost of a run depends on the model, the resolution, the number of images and your plan. It always shows on the Generate button, and next to each option in the model and resolution menus, before you launch anything. See Credits.
See also
- Templates: ready-made recipes
- Video: animate an image
- Nodes: build repeatable workflows
- Gallery: projects, scenes and favorites
Last updated: October 18, 2026