AnkhKit

Open-Source Licenses

AnkhKit builds on outstanding open-source work. This page lists the AI models and key libraries used by our on-device tools. Model weights are downloaded once and cached in your browser from our own servers — but their upstream licenses still apply to the outputs you create.

AI models

Whisper base (OpenAI Whisper)

Used by: Audio transcription — audio-to-text, subtitle generator, speech-to-text, speaking speed test

MIT License Upstream project

Kokoro-82M v1.0

Used by: Text-to-speech synthesis

Apache License 2.0 Upstream project

MODNET

Used by: Background removal — fast mode

Apache License 2.0 Upstream project

BiRefNet_lite

Used by: Background removal — quality mode

MIT License Upstream project

Potrace (via esm-potrace-wasm)

Used by: Image vectorizer

GNU GPL v2 — see packages/tools-core/NOTICE.md for the corresponding source Upstream project

Key libraries

@huggingface/transformers (Transformers.js)

Apache License 2.0 Upstream project

kokoro-js

Apache License 2.0 Upstream project

ffmpeg.wasm

LGPL/GPL (upstream FFmpeg) Upstream project

pdf-lib (@cantoo fork) & pdf.js

MIT / Apache License 2.0 Upstream project

fflate, SheetJS, fast-xml-parser, MarkdownIt & other open-source libraries

MIT / Apache 2.0 (per project) Upstream project

Notes

Whisper, Kokoro, MODNET, and BiRefNet run entirely in your browser after a one-time download — nothing you process is ever uploaded (see our Privacy Policy). The full it-tools attribution notice for ported logic lives in packages/tools-core/NOTICE.md in the open-source repository.