How to check that PixVibe does not upload your files
The free PixVibe tools below process images, PDFs and videos inside your browser. The files you add are read and converted on your device and are not uploaded to PixVibe servers. You do not have to take our word for it: the two checks below work without trusting any PixVibe code.
Check it yourself
1. The offline test
- Open a tool, for example the image compressor, and process one file. The first run downloads the models and codecs the tool needs.
- Turn off Wi-Fi or switch on airplane mode.
- Add another file and process it with the same settings. Processing and downloading the result still work, because nothing has to go to a server.
- The badge in the lower right corner of the tool page counts the files processed while offline.
2. The browser Network panel
- Open developer tools: F12 on Windows and Linux, Cmd + Option + I on a Mac. Choose the Network tab.
- Process a file in the tool.
- The list shows downloads of models and codecs (GET requests) and a few small JSON requests described below. No request carries your file in its body.
Chrome DevTools — Inspect network activity
The badge on each tool page
Every tool listed below shows a badge in the lower right corner with the amount of file data uploaded from the page (0 bytes) and every request the page sent to a server, grouped by type. Open it to see the list while you work.
What runs in your browser
Each tool and the in-browser software that does the work:
Remove BackgroundIS-Net and BiRefNet segmentation models run with onnxruntime-web 1.30.0, on the GPU through WebGPU where available.
Image ConverterjSquash codecs (MozJPEG, OxiPNG, libwebp, libavif) compiled to WebAssembly.
Compress ImagejSquash codecs in WebAssembly and image-q color quantization for smaller PNG files.
Resize Imagepica resampling in the browser, then the jSquash encoders for the output format.
Crop ImageThe browser canvas crops the image, then the jSquash encoders write the file.
AI UpscalerESRGAN models from UpscalerJS, run with TensorFlow.js or onnxruntime-web.
Image to PDFpdf-lib (@cantoo/pdf-lib 2.11.1) builds the PDF in the browser.
PDF to Imagepdf.js 6.4.299 from Mozilla renders each page in the browser.
Compress PDFpdf-lib reads the PDF, the browser re-encodes the JPEG images inside it in a Web Worker, and pdf-lib writes the smaller file. The Smallest level renders pages with pdf.js first.
Passport PhotoYuNet (an open-source face detector from OpenCV) finds the face, MODNet (an open-source portrait matting model) cuts out the person, and the photo is framed, recoloured and laid out for printing on a canvas in your browser.
Compress VideoThe browser WebCodecs VideoEncoder encodes H.264, and mediabunny writes the MP4 file.
What the page does send to PixVibe
None of these requests carries file names or file contents. The badge lists each one as it happens.
Usage statsTool name, settings, number and size of files, processing time, browser language and device type. Used to count how each tool is used.
Analytics (Microsoft Clarity)Microsoft Clarity records clicks, scrolling and the page structure and sends them in compressed form, so we can see where pages are hard to use. The tool workspace is masked in these recordings: file names and file contents are not recorded.
Device infoBrowser, operating system, CPU cores, memory and GPU capabilities, sent once a day per device and again when the device details change. Used to see which devices cannot run the local models.
Error reportsSent only when processing fails: the error message, the file type and size, and the first 16 bytes of the file. The 16 bytes identify the file format (a JPEG starts with FF D8 FF) and do not show the content. Used to collect errors from in-browser processing and fix them.
AccountSign-in status and credit balance, when you are signed in.
When a file is uploaded
- The cloud tools upload the file: AI Background Remover HD · AI Image Enhancer · HD Image Converter run on PixVibe servers and cost credits.
- Cloud refine on the background remover opens the cloud background remover, and the badge switches to say that this step uploads.
- Uploaded files and results are deleted within 24 hours, as stated in the privacy policy.
What the badge cannot see
- The badge records requests made by the page itself. Model downloads inside Web Workers are not listed there; they are downloads (GET), and the Network panel shows them.
- For an independent check, use the offline test or the Network panel above.