Open-source licenses
Eziclip.com runs entirely on open-source libraries and AI models, right in your browser. Here's what powers it and where to find the source.
Video & audio
Video, GIF and audio processing use FFmpeg compiled to WebAssembly (ffmpeg.wasm). The binary we ship is licensed under the GPL-2.0-or-later. The corresponding source and our build configuration are available upstream at github.com/ffmpegwasm/ffmpeg.wasm.
AI models
Captions use OpenAI Whisper (MIT) via Transformers.js (Apache-2.0). Translation uses Helsinki-NLP Opus-MT (Apache-2.0). Background removal uses BiRefNet (MIT). Upscaling uses Real-ESRGAN (BSD-3). Object removal uses LaMa (Apache-2.0). OCR uses Tesseract (Apache-2.0). Voice noise removal uses RNNoise (xiph/rnnoise, BSD-3-Clause) via @shiguredo/rnnoise-wasm (Apache-2.0). Studio and Text to Speech voiceovers use Kokoro-82M (Apache-2.0) for natural English, with a non-GPL English grapheme→phoneme step built on the CMU Pronouncing Dictionary (BSD-2-Clause) via cmu-pronouncing-dictionary (ISC), and Supertonic 3 (OpenRAIL-M model weights; sample code MIT) for plain multilingual readers. They run in your browser. OpenRAIL-M restricts certain uses of the model weights — see the license on the model card when redistributing those weights.
Image codecs
Image conversion uses jSquash (Apache-2.0). HEIC decoding uses libheif via heic-to (LGPL-3.0), loaded as a separate module.
A full machine-readable inventory is maintained in the project's NOTICE.md.