Why the batch is where privacy actually bites
Doing one image on an upload-based site is a small exposure. Doing fifty is a folder — and folders are personal in a way single files are not. A batch is usually an event, a shoot, a set of documents, a holiday. It is the whole context, not one frame of it.
Nothing here is uploaded, so the batch stays a batch on your own machine. Every image is also stripped of its metadata on the way through, which for a folder of phone photos means fifty sets of GPS coordinates that do not travel with the files.
How the work is paced
Images are processed one at a time, and each canvas is released before the next begins. That keeps memory flat regardless of how many files you add, rather than climbing until the tab is killed — which matters on a phone, where the browser will discard the page without much warning.
Progress is shown per image and you can cancel at any point. Cancelling leaves nothing behind; the ZIP is only assembled once every image is done.
Fifty is the cap, and it is a practical rather than an arbitrary one. Beyond that the wait stops feeling interactive and a desktop tool is genuinely the better answer.
Sensible settings for a batch
Setting the longest edge rather than exact dimensions is what makes one setting work across a mixed batch: portrait and landscape images both end up correctly constrained without you sorting them first. 1600 pixels is a good general web default; 2400 if the images will be viewed full screen.
Images already smaller than the box are passed through without enlarging, so mixing a few small graphics into a batch of photos does no damage.
Converting a batch to WebP typically saves 25 to 35% over JPG at matching quality. If the images are going on a website, that is the single easiest win available.
What a batch tool deliberately does not do
It applies one setting to everything. That is the entire value — and the entire limitation. If half your images need a different treatment, they are two batches, not one run with exceptions. Trying to encode per-image rules into a bulk tool produces an interface nobody can predict.
It also does not crop, reframe or hit a target file size per image. Cropping needs a decision about what to keep, which is different for every picture; a per-image size target needs a search per image and turns a fast batch into a slow one. Those live in the single-image tools, where you can see what you are doing.
What it does do is the boring, repetitive part: make fifty photos a sensible size and a sensible format, keep their names, and hand them back in one file. That is usually the whole job.