Step by step: product photos edited with GIMP and an AI agent

Image editing has two halves: the bulk work, boring and repetitive, and the fine retouching, which needs a human eye. On an AgentBox you split them properly: the agent processes hundreds of files from the terminal, and you open in GIMP only the photos worth the attention. This guide uses a real scenario: product photos for an online shop.

Step 1: install the image editor

On the AgentBox tab, under Apps, click Install on Image editor. The box gets GIMP, the established open source photo editor, on the desktop you reach from any browser.

Step 2: get the photos up

Drag the files into the console's Files view or hand the agent a download link. For a phone full of product shots, the simplest route is a shared folder or an archive: the agent downloads and unpacks it by itself.

Step 3: the bulk work, in the terminal

The Poze folder holds the raw product photos. Crop them square on center, resize to 1600 by 1600, balance the brightness, save as web-optimized JPEG and rename them after the list in products.csv.

The agent writes and runs the processing scripts and shows you a few samples. If the background needs whitening or the colors need adjusting, you say so on the samples and it redoes the whole batch with your corrections. A hundred photos take minutes, not an afternoon.

Step 4: the fine retouching, in GIMP

The hero shots (each product's first image, the campaign banner) deserve a human hand. Switch to the Desktop view, open the file in GIMP and work the classic way: cloning, levels, careful crops. The agent can open the right file for you: open product X's main photo in GIMP.

Step 5: export and handover

At the end, the agent packages the result: the JPEG folder for the site, the thumbnails, the archive to deliver. If the shop has an API or FTP, it can upload them too, and the whole operation becomes repeatable for every new batch of products. For graphics drawn from scratch rather than from photos, see the Krita guide and the logo in Inkscape.

What you end up with

• Whole batches of images processed by the agent to your rules.
• GIMP on the live desktop for the touches that need a human eye.
• A repeatable flow for every new batch of photos.