Third-party components¶
| Component | Use | License |
|---|---|---|
TIGER (look2hear) code, vendored and trimmed in dubber/third_party/look2hear |
dialogue / effects / music separation | MIT (see dubber/third_party/look2hear/LICENSE) |
TIGER-DnR weights (JusperLee/TIGER-DnR) |
downloaded on demand | Apache-2.0 |
| faster-qwen3-tts, qwen-tts-hf | Qwen3-TTS inference, CUDA Graphs | MIT / Apache-2.0 |
| Qwen3-TTS-12Hz Base models (Qwen) | TTS with voice cloning, downloaded on demand | Apache-2.0 |
faster-whisper, CTranslate2, deepdml/faster-whisper-large-v3-turbo-ct2 |
speech recognition | MIT |
| Silero VAD | voice activity detection | MIT |
Opus-MT tc-big (Helsinki-NLP/opus-mt-tc-big-en-zle, -zle-en, -zle-de, -de-zle) |
translation (en↔ru, ru↔de) | CC-BY-4.0. University of Helsinki (Helsinki-NLP), https://huggingface.co/Helsinki-NLP, https://creativecommons.org/licenses/by/4.0/ |
Opus-MT tc-bible-big (Helsinki-NLP/opus-mt-tc-bible-big-deu_eng_fra_por_spa-gmw, Helsinki-NLP/opus-mt-tc-bible-big-gmw-deu_eng_fra_por_spa) |
translation (en→de with >>deu<<, de→en with >>eng<<) |
Apache-2.0. University of Helsinki (Helsinki-NLP), https://huggingface.co/Helsinki-NLP, https://www.apache.org/licenses/LICENSE-2.0 |
| pyannote speaker-diarization community-1 | speaker diarization | CC-BY-4.0. Hervé Bredin and the pyannote.audio authors. Original https://huggingface.co/pyannote/speaker-diarization-community-1 . Mirror https://huggingface.co/Mitroshenkov87/voxprint-pyannote-speaker-diarization-community-1 at 9c851af727f1085a361777d13e32734f7f5ebb30. https://creativecommons.org/licenses/by/4.0/ . No changes besides removing the access gate. |
| pyannote/embedding | speaker embeddings | MIT. Copyright (c) 2022 CNRS. Original https://huggingface.co/pyannote/embedding . Mirror https://huggingface.co/Mitroshenkov87/voxprint-pyannote-embedding at 4655ca656e1d7b1d8fe8100b39b4fdbf3015c7b8. No changes besides removing the access gate. |
Mel-Band RoFormer (KimberleyJSN/melbandroformer @ ac9b0614ab3cd7f77219e18ba494dfd93956c348, 2026-04-22) |
optional vocal separation | MIT (model card relicensed MIT on that commit) |
| PySide6 / Qt | user interface | LGPL-3.0 |
| FFmpeg libraries inside PySide6 (Qt Multimedia FFmpeg backend, shared DLLs) | built-in video player for the preview and Watch mode | LGPL-2.1-or-later (dynamically linked, replaceable) |
| python-stretch (Signalsmith Stretch) | time stretching of dub lines | MIT |
| PEFT | loading voices of the shared Voxprint voice library (LoRA adapters) | Apache-2.0 |
| SubDL / OpenSubtitles APIs | optional subtitle search with the user's own keys | service terms of each site |
| PyTorch, transformers, numpy, scipy, librosa, soundfile | runtime | BSD / Apache-2.0 / ISC |
| ffmpeg program (an LGPL build is preferred; PATH or imageio-ffmpeg as a fallback) | audio extraction and muxing, called as a separate program | LGPL/GPL, depending on the build |
espeak-ng (only to generate the synthetic test clip, tools/make_test_clip.py) |
not shipped | GPL-3.0 (the generated audio is synthetic) |
Project licence: Apache-2.0 for the program's own source code (see LICENSE and NOTICE), the same as Voxprint AI Audiobook Builder.
dubber/assets/splash-easter.jpg is original generated art shipped with this program. It does not depict a real person. See dubber/assets/README.md.