Pixmunk
Tools Remove background Blur background Upscale image Colorize a photo Remove an object Compress images Image to text (OCR) Passport photo Enhance a photo
Guides How to Remove a Background How to Upscale an Image How to Restore an Old Photo What WebGPU Means for Image Tools The Transparent PNG Guide
Privacy How Private Is This? About Open Source Licenses

Open source licenses

By Michael Machatschek Updated

The tools on this site run inside your browser, and much of what runs there is open-source software written by other people. This page credits it and meets the license obligations that come with it. None of these components ever receive your images; they are the code your browser runs locally.

Libraries loaded at runtime

libheif and libde265, via the heic-to package. These decode HEIC photos in the image compressor and HEIC converter. Both are licensed under the GNU Lesser General Public License v3.0 (LGPL-3.0). As the LGPL requires, we load them as separate, unmodified files at runtime rather than bundling them into our own code, so you (or we) can swap in your own build of the library. Source code: libheif, libde265, heic-to. The full license text is at gnu.org/licenses/lgpl-3.0.html.

jSquash (Apache-2.0), WebAssembly builds of the image codecs mozjpeg (IJG, BSD-3-Clause, zlib), libwebp (BSD-3-Clause) and libavif with libaom (both BSD-2-Clause, with the Alliance for Open Media Patent License 1.0). Used where a browser cannot encode a format itself. Source: jSquash.

Tesseract.js and tesseract-core (Apache-2.0), the OCR engine behind image to text, including its Apache-2.0 language data. Source: tesseract.js, tesseract-ocr.

MediaPipe Tasks (Apache-2.0), Google's face landmark detection, used by the passport photo maker to measure head position. Source: mediapipe.

ONNX Runtime Web (MIT) and Transformers.js (Apache-2.0), the two inference runtimes every AI tool here runs on. Source: onnxruntime, transformers.js.

A note on the AI models

The neural networks themselves are all under permissive licenses that allow commercial use (Apache-2.0, MIT, BSD). We verify this before shipping any model, and re-verify before swapping one, because several well-known models in these categories are research-only and may not be used on a commercial site. We do not publish which specific checkpoints we use.