How it works
Choose a task, adjust the settings, review the result and download your finished file.
Four steps
Pick a guided task or a single tool, then choose your files. Each file is identified from its actual structure, not just its extension, so you know whether it can be processed, only inspected, or is unsupported before anything changes.
Set available dimensions, compression, supported metadata fields or a watermark in the appropriate tool. Settings only describe the job; nothing runs until you start it yourself.
Compare original and result: check the visible image, dimensions, file size and available metadata information. Completed guided versions show supported fields added, changed, removed, retained, not found in either checked file, or not checked. These are bounded checks, not an absence guarantee. Measured sizes can grow as well as shrink.
Save the finished copy. Your original stays unchanged, and results live only in this tab’s memory, so download what you want to keep.
Real sample
This is the actual demo: a generated fictional file, processed on this device only when you press the button. Checks shown after running are computed on the real output.
Original: fixed synthetic example, nothing processed.
This explicitly prepares app resources if needed and cleans only the chosen fixed fictional example. Nothing is uploaded. Your own files still need Pro to clean.
Metadata lists are read from the actual supported text/EXIF fields before and after. Each check is reported separately and “Not checked” is never treated as verified.
These tiny generated examples contain no location or personal photo. They do not demonstrate GPS, PDF or Office cleaning. Identifying visible content is not redacted.
Guided tasks
Crop, resize, choose a format and compression setting, and optionally place a PNG logo, for up to 10 JPEG, PNG or WebP images. Background removal is a separate single-image tool that runs a local AI model; it is never part of the guided batch. A visible text watermark is also its own tool.
Open the webshop taskInspect the readable fields in JPEG and PNG photos, then edit or remove supported fields, or only inspect. Look at the picture itself before sharing: metadata removal does not hide faces, signs or visible text, and it is not redaction.
Open the share photos taskFor an unencrypted, supported PDF: inspect, edit or clean its supported properties, review the page preview, and save the checked PDF. Export only the embedded assets actually available, and extract the text layer on request if present. There is no OCR and no page organizer.
Open the client document taskFile contents, names, metadata, previews and outputs remain in browser memory on this device. Originals are never changed; each result is a new copy held in this tab’s memory, not a durable cloud history.
Loading the site needs a connection initially. Offline app resources and selected AI models and runtimes need explicit online preparation; models are not downloaded just by opening the homepage. License activation and recovery services, payments, updates and opening external support links use a connection separately from your files. Current signed recovery files can restore locally; legacy license upgrades need a connection. PixelRinse has no application usage telemetry. Hosting and payment providers may keep operational records independently of local file processing.
Read the privacy pageIf the browser clears stored data to free space, you may need to prepare resources again.
Inspecting supported readable files and running the exact built-in demos is free. Processing your own files requires the existing Individual Pro license.
No. File contents, names, metadata, previews and outputs stay in this browser’s memory on your device. A connection is only used for app resources, optional model downloads and license, payment, update or support links, never to send your files.
Guided tasks: webshop accepts JPEG, PNG and WebP; share photos accepts JPEG and PNG; the client document task accepts unencrypted supported PDFs. Guided tasks take up to 10 files, 25 MiB per file and 200 MiB combined. The ordinary cleaner accepts up to 100 unread file references / 1 GiB, with active groups of 20 files, 25 MiB each and 200 MiB total. Other tools have their own safety, pixel, page and runtime limits; check the tools directory and file support page.
Yes, once this build’s application resources have been prepared while online, plus the selected model and runtime for tools that need one, and a valid local license for processing your own files. If the browser evicts stored resources, you may need to prepare them again.
Nothing. PixelRinse always writes a new copy. Results live in this tab’s memory, not a cloud history, so download what you want to keep before closing the tab.
Supported embedded fields for the detected format, listed per file as removed, retained or not checked. It does not remove faces, visible text, landmarks or anything else in the pixels or page content.
Yes, within the limits above. Each file gets its own outcome, so an unsupported or failed file never hides the others. Background removal is single-image only and is not part of a batch.
Yes. Inspecting supported readable files is free, and the built-in demos, including the sample on this page, run exactly as shown without a license. Processing your own files requires Individual Pro.
Read the file’s refusal or failure reason; it is not offered as a successfully processed output. Check the format and tool limits, try a smaller supported copy or another appropriate tool, and keep your original. Encrypted, malformed, ambiguous or out-of-scope files are refused rather than receiving a cleaned label. Help has troubleshooting steps; do not send private files to support.
More answers in the full FAQ.