PyroSoft Convert - https://pyrosoft.pro/convert/
Copyright 2026 PyroSoft Productions, Inc.

PyroSoft Convert runs entirely in your browser and is built on these open-source projects.
Their unmodified distribution files are served from /convert/vendor/. Each project's license text
is in /convert/vendor/licenses/ where the project ships one. The GPL and the font licenses are in /convert/assets/licenses/.

FFmpeg (via ffmpeg.wasm @ffmpeg/core and @ffmpeg/core-mt 0.12.10)
  License: GPL-2.0-or-later (the WebAssembly build includes GPL components such as x264 and x265). Full text: /convert/assets/licenses/GPL-2.0.txt
  Source: hosted here in /convert/vendor/source/ffmpeg/ (FFmpeg 5.1.4, x264, x265 and the other linked libraries, plus the ffmpeg.wasm
  build scripts). Exact versions and commits are in /convert/vendor/source/README.txt, and checksums in /convert/vendor/source/SHA256SUMS.
  The .wasm files are served gzip-compressed and are otherwise unmodified.
ffmpeg.wasm @ffmpeg/ffmpeg 0.12.15 and @ffmpeg/util 0.12.2 - MIT - https://github.com/ffmpegwasm/ffmpeg.wasm
Mediabunny 1.61.4 - MPL-2.0 - https://github.com/Vanilagy/mediabunny (unmodified)
three.js r186 (0.186.1) - MIT - https://github.com/mrdoob/three.js
  Includes the Draco decoder (Apache-2.0), Basis Universal transcoder (Apache-2.0), meshoptimizer decoder (MIT),
  chevrotain (Apache-2.0), fflate (MIT), ktx-parse (MIT) and zstddec (MIT) as distributed with three.js.
pdf.js 6.4.299 (pdfjs-dist) - Apache-2.0 - https://github.com/mozilla/pdf.js
pdf-lib 1.17.1 - MIT - https://github.com/Hopding/pdf-lib
heic-to 1.6.6 - LGPL-3.0 (includes libheif / libde265, LGPL-3.0) - https://github.com/hoppergee/heic-to
  The library is served unmodified as a separate file and may be replaced with any compatible version.
jSquash AVIF encoder 2.1.1 (libavif / libaom) - Apache-2.0 and BSD-2-Clause - https://github.com/jamsinclair/jSquash
fflate 0.8.3 - MIT - https://github.com/101arrowz/fflate
js-yaml 5.4.3 - MIT - https://github.com/nodeca/js-yaml
smol-toml 1.9.1 - BSD-3-Clause - https://github.com/squirrelchat/smol-toml
fast-xml-parser 5.11.2 - MIT - https://github.com/NaturalIntelligence/fast-xml-parser
Papa Parse 5.7.0 - MIT - https://github.com/mholt/PapaParse
marked 18.1.1 - MIT - https://github.com/markedjs/marked
Inter (font) - SIL Open Font License 1.1 - https://github.com/rsms/inter
JetBrains Mono (font) - SIL Open Font License 1.1 - https://github.com/JetBrains/JetBrainsMono

AI models and runtimes for the Audio tab. Weights are served unmodified from /convert/models/; the runtimes from /convert/vendor/ai/.
ONNX Runtime Web 1.31.0-dev - MIT - https://github.com/microsoft/onnxruntime
Transformers.js 4.3.2 - Apache-2.0 - https://github.com/huggingface/transformers.js (import paths changed to load from this site)
Whisper (tiny, base, small, large-v3-turbo; OpenAI), ONNX exports by onnx-community - MIT - https://github.com/openai/whisper
HTDemucs (Meta AI), 21-piece ONNX export by monteslu - MIT - https://github.com/facebookresearch/demucs and https://huggingface.co/monteslu/htdemucs-web-onnx
demucs-web signal processing - MIT - https://github.com/timcsy/demucs-web (the model wrapper was removed)
DeepFilterNet 3 (Hendrik Schroeter), ONNX export by kimtos-labs - MIT or Apache-2.0 - https://github.com/Rikorose/DeepFilterNet
Kokoro-82M v1.0 (hexgrad), ONNX export by onnx-community - Apache-2.0 - https://huggingface.co/hexgrad/Kokoro-82M
kokoro-js 1.2.1 - Apache-2.0 - https://github.com/hexgrad/kokoro (import paths and the voice location changed to load from this site)
phonemizer 1.2.1 - Apache-2.0 - https://github.com/xenova/phonemizer.js, which includes eSpeak NG compiled to WebAssembly - GPL-3.0-or-later. Full text: /convert/assets/licenses/GPL-3.0.txt
  Source: hosted here in /convert/vendor/source/espeak/ (phonemizer 1.2.1 and eSpeak NG). /convert/vendor/source/README.txt explains how the
  eSpeak NG commit was identified, since phonemizer ships it precompiled without naming one, and what could not be matched exactly.
  It is served unmodified as a separate file and is used only for English text to speech.
Supertonic 3 (Supertone Inc.): code MIT - https://github.com/supertone-inc/supertonic
  Model weights: OpenRAIL-M - https://huggingface.co/Supertone/supertonic-3/blob/main/LICENSE
  The OpenRAIL-M license restricts how the model and its output may be used, and those restrictions apply to you when you use it here.
  Among other things, it may not be used to break the law, to harm or exploit people or minors, to impersonate someone or
  generate speech in a real person's voice without consent, to spread disinformation meant to harm, or to discriminate. Read the license for the full list.