Image groups and batch processing
Collect product photos into an image group, run one prompt over every picture, name the files and download them as a zip.
An image group holds several images in one card. Wire a group into an image prompt and Canvas renders once for every picture in it — the quick way to swap backgrounds, match a style or produce product shots in bulk.
Make an image group
- On the board: select two or more finished images (drag a selection box, or ⌘ / Ctrl-click them), then choose Group into one in the bar at the bottom. Pictures are ordered top to bottom, then left to right, and duplicates of the same image are kept once. ⌘ / Ctrl + Z undoes it.
- Add to an existing group: select a group together with other images and the bar offers Add to group.
- Automatically: picking several Shopify product images at once, or sending several photos from your phone, arrives as a group.
New groups are named "Group 1", "Group 2" and so on; click the title on the card to rename it.
The group card
| Action | What it does |
|---|---|
| Open all | Shows every picture; double-clicking the thumbnails does the same |
| Download zip | Downloads the finished pictures as one zip file |
| Ungroup | Turns the finished pictures back into single image cards; unfinished ones are dropped |
| Remove | Deletes the group card |
The card's footer shows how many pictures it holds, and "N/M done" while a batch is running. If some pictures fail, "N failed — retry" re-runs just those, with the upstream prompt card's current settings. The group has to still be wired to its prompt card to retry.
See every picture
In Open all:
- Edit the file name each picture will download with, under the picture.
- Remove from group takes a picture out of the group; it stays in your Assets library.
- Click a picture to enlarge it, and use ← → or the previous / next buttons to move between them. Batch results show side by side with the photo they came from, labelled source and result.
Run a batch
Wire the group into an image prompt card. The card shows "Runs over N images".
Write the prompt. Describe the treatment, not the product — for example "white studio background, soft shadow, product centred". Each result keeps its own product.
For a consistent look, wire in one more single image as a style reference; it's used for every picture.
Check the estimate for the whole run on the button (pictures × images per picture), then generate.
All the results are collected in a new image group, wired after the prompt card. Every picture renders on its own, so one failure doesn't stop the others.
With a group wired in, an image prompt card takes up to 3 more references, because every picture in the group is itself image 1.
Name the downloaded files
When a group is wired in, the prompt card shows File names. Left empty, results keep the source photo's name. You can use these variables:
| Variable | Meaning |
|---|---|
{board} | The board's name |
{index} | A three-digit number, such as 001 |
{source} | The source photo's file name (without the extension) |
{ai:…} | Has AI name the picture from what it shows; write your instruction after the colon |
For example, {board}_{ai:a 2-4 word product name, no brand}_{index} gives names like summer-launch_tote-bag_001. With the field empty, Name them with AI fills in that example. Unknown variables are flagged in red.
AI naming adds a small AI charge. If naming fails, the file name keeps its other parts.
Download as a zip
Download zip on the group card packs the finished pictures into one zip, named after the group. Each file takes the name you edited, otherwise the original name, otherwise its number; duplicate names get a suffix such as _2.