Preview
Last updated 28 August 2026
Why one at a time
It is tempting to fire every image through the model at once, and it is the quickest way to crash a browser tab. Each image being processed holds its pixel data, the model's working memory and the result all at the same time, and a browser gives a page far less memory than a desktop application. Ten images in parallel means ten of everything, which on a phone runs out long before it finishes. Processing sequentially takes the same total time on most machines — the work is the same, just queued — and it keeps peak memory at roughly one image rather than all of them.
The quality setting is the other lever. Standard mode caps the long side at 1500 pixels before processing, which is plenty for anything going on a website or a marketplace listing and is dramatically faster on a phone. Original keeps the full resolution, which matters for print or when you are cutting out a photo you will crop into later. If a batch is stalling, drop to standard first — it usually solves it.
Common questions
How many images can I do at once?
Twenty per batch. The limit is about memory rather than time — beyond that, browsers on phones tend to run out and kill the tab partway through, which loses everything. Run a second batch instead.
What kind of images work best?
Photographs of people. The model was trained mainly on human subjects, so portraits and full-body shots come out cleanest. Products, objects and animals still work, but edges are less reliable — especially around fine detail like hair, fur and thin straps.
What is the difference between the quality settings?
Standard caps the long side at 1500 pixels before processing, which is faster and fine for web use. Original keeps the full resolution for print or heavy cropping, at the cost of noticeably longer processing and much higher memory use.
Are my images uploaded anywhere?
No. Every image is read, processed and zipped entirely on your device. What is downloaded is the AI model, once, from public open-source CDNs — the photographs never leave your machine.
Why did one image fail while the rest worked?
Usually memory, or an unusual format the browser could not decode. Failed images are marked in the grid and skipped rather than stopping the batch, so you still get everything else. Retry the failed one on its own, in standard quality.