Processing one image is easy. Processing a thousand is a project. The difference is not just the time. It is the workflow, the quality control, and the failure handling.
Batch processing makes sense when you have more than about fifty images and the settings are consistent across all of them. If the settings vary, individual processing is faster.
For ecommerce catalogs, the settings are usually consistent. Same background colour, same model, same resolution. This is the ideal case for batch.
For mixed content (portraits and products in the same batch), the router on Remova handles the variation automatically. You can still batch but the processing time varies per image.
Group your images by category. All product photos together, all portraits together, all logos together. This lets you apply consistent settings per group.
Check for obvious issues before processing. Blurry images, tiny images, images with no clear subject. These will fail or produce poor results regardless of the tool.
Compress oversized files before uploading. A 20MB image takes longer to upload than a 3MB image, and the extra resolution rarely helps the segmentation quality.
The batch endpoint on Remova accepts up to twenty images per request. For a catalog of two hundred, that is ten requests. For two thousand, a hundred requests.
The rate limit is thirty requests per minute per IP. At twenty images per request, that is six hundred images per minute. The rate limit is not the bottleneck for most catalogs. The processing time is.
Each batch takes roughly twenty seconds per image on the server side. A batch of twenty takes four hundred seconds, or about seven minutes. Plan accordingly.
Sample the output. Look at ten percent of the results. If the failure rate is under five percent, proceed with the full catalog. If it is higher, adjust the source or the settings.
Common failures: images where the subject blends into the background, images with transparent objects, images with very fine details. These are the same categories where single-image processing also struggles.
Reprocess failures individually. Some will succeed with a different model. Some need manual editing. Some should be reshot.
Keep originals separate from processed images. Use folder structures like input/product-name/ and output/product-name/ so the relationship is clear.
Name processed files systematically. original-filename-processed.png makes it easy to match processed results back to originals.
Keep a spreadsheet of which files were processed with which settings. Useful if you need to reprocess with different settings later.
For catalogs over ten thousand images, split the work across multiple sessions. There is no point processing everything at once.
If you are a developer, use the API directly and log results to a database. This lets you track success rates over time and reprocess failures systematically.
Consider running the process overnight. The rate limit resets and the processing continues. You wake up with a catalog of cutouts.
Twenty images per request. Larger catalogs are processed as multiple requests.
No. Video uses a separate pipeline. Each video is processed individually. Ten videos per batch is the practical limit.
The API returns a status code per request. Log the failures and reprocess them separately.
Yes, on the client side. The actual server-side processing time per image is the same. The savings come from less manual clicking and fewer round trips.