Lower the quality, cap the dimensions, or convert to WebP in one click for maximum compression. Batch processing tells you how much each image saved.
Will compression be visible?
Between 70–85 quality the difference is usually invisible while cutting file size to 30–60%. Try WebP for even smaller files.
Are my images uploaded?
No. Everything runs in your browser; files never touch a server.
Everything runs locally in your browser — files are never uploaded. BoxKit VIP unlocks unlimited use of all 29 tools.
Look at the other side's limit first, then decide how far to compress. Form uploads often cap a single image at 2MB, email attachments are commonly 20MB, e-commerce main images are mostly required to be under 500KB, and inline images in WeChat articles over 1MB get recompressed by the platform. In these cases do not agonise over image quality; getting under the line is enough, and quality around 82 is fine.
Web images are another category. A phone photo straight out of the camera is often 3000–4000 pixels wide and 3–5MB, and dropping it into a page slows the first screen. Such images should be scaled down to the actual display width first and then compressed, not compressed hard at the original size.
Three cases are unsuitable: an image that has already been compressed once (a second compression keeps damaging what is already damaged, and detail does not come back); scans and documents for OCR or printing (lossy compression softens stroke edges); and original shots that need long-term archiving. Keeping the original file is the better deal in these cases.
The difference in online compression services is mainly privacy and quota: the file has to be uploaded to their server, there may be a queue and a limit on the number of images, and once the free quota runs out you pay. Here everything is done locally in your browser and files are never uploaded, so there is no per-file size limit and no usage limit, and it works offline. The cost is that speed depends on your own device — processing dozens of 40-megapixel photos in one go on a phone gets noticeably slow or freezes — and the result is gone once you close the page, with no cloud record.
Compared with Photoshop's Save for Web, there is no selective region compression and no post-compression sharpening compensation here, only the two knobs of quality and edge length. For per-image fine-tuning you still need desktop software.
Compared with tools such as ImageOptim and pngquant, which are stronger at lossless optimisation of PNG and do palette quantisation, this tool can basically only keep a PNG as it is or change the format. It is good enough for photos, but it has no advantage on PNG icons.
Why does the file size jump so much going from quality 90 to 95?
JPEG file size is not linear in quality. Above quality 85, the quantisation table cuts high-frequency coefficients less and less, so the bits needed to encode each 8×8 block rise quickly. In practice, going from quality 85 to 100 can grow the file three to five times while the visible difference is tiny. So 100 does not mean best, only biggest.
Why does compressing below 100KB make it blurry?
A file-size cap is a fixed bit budget for the whole image. Compressing a 1200×1600 image to 100KB leaves about 0.04 bit per pixel on average. The compressor can only prioritise low-frequency information (the general outline and colour blocks), so high-frequency texture, hair and stroke edges are largely discarded, and bands appear in the sky while faces look smeared. If you need it smaller, the right move is to reduce the pixel count, not to keep lowering quality.
Is WebP always better than JPEG at the same file size?
Most photos are, especially below about 150KB, where WebP has less blocky noise and cleaner gradients and text edges; in the high-quality range the gap narrows until it is invisible. The exception is flat-colour icons and pixel art, where lossless PNG is actually clearer. Also, converting to WebP re-encodes, so EXIF (capture time, camera model, GPS) is lost.
Is it useful to compress an already compressed image again?
Little point, and harmful. The detail lost in the first compression does not come back; the second only deepens block boundaries on the existing blocking, and the file size usually does not drop much and sometimes grows (the result will be marked as +X%, in which case keep the original). To recompress, go back to the most original version.