Open Source and Attributions
The open-source audio libraries Rifflo delivers to your browser, their licences and source, the fonts and icons we use, and the demo track credit.
Last updated: 2026-11-02
Rifflo runs its audio engine inside your browser, which means some open-source components are delivered to your device as part of the application. This page lists them with their licences and source locations, together with the fonts, icons, template and demo material we use. Licences were verified by reading the licence files inside each package, not only the metadata.
1. Audio libraries delivered to your browser
| Package | Version | Licence | Source | Used for |
|---|---|---|---|---|
| mediabunny | 1.65.0 | MPL-2.0 | github.com/Vanilagy/mediabunny | Reading MP4/M4A, MP3, Ogg and ADTS containers (sample rate, channels, duration, decoder configuration); writing the FLAC container on export |
| @mediabunny/flac-encoder | 1.65.0 | MPL-2.0, wrapping libFLAC 1.5.x (Xiph.Org BSD-style licence, reproduced in section 2) | github.com/Vanilagy/mediabunny/tree/main/packages/flac-encoder ; libFLAC: github.com/xiph/flac | FLAC export at 16 and 24 bit, inside the engine worker |
| idb | 8.0.4 | ISC. Copyright (c) 2016, Jake Archibald | github.com/jakearchibald/idb | Promise wrapper for IndexedDB (project metadata and the IndexedDB fallback of the file store) |
| @mediabunny/mp3-encoder | 1.65.0 | MPL-2.0, wrapping LAME 3.100 (LGPL 2.1, see section 3) | github.com/Vanilagy/mediabunny/tree/main/packages/mp3-encoder ; LAME: lame.sourceforge.io | MP3 export in the studio, the converter and the cutter. Delivered as a separate file, loaded the first time you export an MP3 |
| music-metadata | 12.3.4 | MIT. Copyright (c) 2017 Borewit | github.com/Borewit/music-metadata | Reading tags and cover art in the metadata editor and the converter |
| fflate | 0.8.3 | MIT. Copyright (c) 2023 Arjun Barrett | github.com/101arrowz/fflate | ZIP download of batch results in the converter and the metadata editor, and of all stems in the Stem Split Studio |
| signalsmith-stretch | 1.3.2 | MIT. Copyright (c) 2022 Geraint Luff / Signalsmith Audio Ltd. | github.com/Signalsmith-Audio/signalsmith-stretch | Pitch changer and the "tempo only" option of the nightcore maker: time-stretch and pitch-shift, run offline in a worker |
| @spotify/basic-pitch (code and model weights) | 1.0.1 | Apache License 2.0. Copyright 2022 Spotify AB | github.com/spotify/basic-pitch-ts | Audio to MIDI: the note transcription model, loaded on that page only |
@tensorflow/tfjs and @tensorflow/tfjs-backend-wasm (with @tonejs/midi 2.0.28, MIT; long 4.0.0, Apache 2.0; seedrandom 3.0.5, MIT) | 3.21.0 | Apache License 2.0. Copyright Google LLC. The WebAssembly backend embeds XNNPACK and pthreadpool (BSD), notices in section 2 | github.com/tensorflow/tfjs | Runs the Basic Pitch model in your browser (WebGL, WebAssembly or CPU), loaded on the audio-to-MIDI page only |
| onnxruntime-web (WebAssembly build) | 1.20.1 | MIT License. Copyright Microsoft Corporation. Third-party notices for code compiled into the WebAssembly file at github.com/microsoft/onnxruntime (ThirdPartyNotices.txt) | github.com/microsoft/onnxruntime | Runs the beat model in your browser, loaded on the key and BPM page (and for the tempo written into a MIDI file) |
Beat This! small model (code and weights, exported to ONNX and quantised by us) | small0 | MIT License. Copyright 2024 Institute of Computational Perception, JKU Linz | github.com/CPJKU/beat_this | Finds beats, downbeats and tempo in your browser |
Respiro-en breath detector (weights exported to ONNX and quantised by us) | 2024-09-18 | MIT License. Copyright (c) 2024 DongYANG. Trained on LibriTTS-R (Koizumi et al.), CC BY 4.0 | github.com/ydqmkkx/Respiro-en | Finds breaths in the breath remover, in your browser |
The WAV, AIFF and RF64 parsers and writers, the FLAC decoder, the sample-rate converter, the ID3v2.4, FLAC, WAV and AIFF tag writers, the BS.1770 loudness meter, the mastering DSP kernel, the beat and key analysis, the reverb, the varispeed resampler, the device-simulation curves and the MIDI file writer are Rifflo's own code and are not open source. Decoding of lossy formats (MP3, AAC, Vorbis, Opus) and AAC and Opus encoding use the WebCodecs and Web Audio APIs built into your browser; those components belong to the browser vendor and are not distributed by us.
2. Notices
Mozilla Public License 2.0 (mediabunny, @mediabunny/flac-encoder)
Rifflo uses mediabunny and @mediabunny/flac-encoder without modification. Under section 3.2 of the Mozilla Public License 2.0 you are entitled to the source code of the covered files, which is available at github.com/Vanilagy/mediabunny. The licence text is at mozilla.org/MPL/2.0. If we ever modify a covered file, we will publish the modified source at the same location and say so here.
libFLAC (inside @mediabunny/flac-encoder)
The FLAC encoder embeds libFLAC from the Xiph.Org Foundation, which requires this notice to be reproduced with binary distributions:
Copyright (C) 2000-2009 Josh Coalson
Copyright (C) 2011-2025 Xiph.Org Foundation
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions
are met:
- Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
- Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
- Neither the name of the Xiph.Org Foundation nor the names of its
contributors may be used to endorse or promote products derived from
this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR
CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
signalsmith-stretch, @tonejs/midi, seedrandom
MIT License. signalsmith-stretch: Copyright (c) 2022 Geraint Luff / Signalsmith Audio Ltd. @tonejs/midi: Copyright (c) 2016 Yotam Mann. seedrandom: Copyright 2019 David Bau.
Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
onnxruntime-web, Beat This!
MIT License. onnxruntime-web: Copyright (c) Microsoft Corporation. Beat This!: Copyright (c) 2024 Institute of Computational Perception, JKU Linz, Austria. Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions: The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software. THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
@spotify/basic-pitch, @tensorflow/tfjs, @tensorflow/tfjs-backend-wasm, long
Licensed under the Apache License, Version 2.0, available at apache.org/licenses/LICENSE-2.0. Distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND. basic-pitch: Copyright 2022 Spotify AB. TensorFlow.js: Copyright 2017 Google LLC. long: by Daniel Wirtz.
XNNPACK and pthreadpool (inside the TensorFlow.js WebAssembly backend)
BSD licences, full texts at github.com/google/XNNPACK/blob/master/LICENSE and github.com/Maratyszcza/pthreadpool/blob/master/LICENSE. XNNPACK: Copyright (c) Facebook, Inc. and its affiliates. All rights reserved. Copyright 2019 Google LLC. pthreadpool: Copyright 2019 Google LLC. Copyright (c) 2017 Facebook Inc. Copyright (c) 2015-2017 Georgia Institute of Technology.
idb
ISC License. Copyright (c) 2016, Jake Archibald. Permission to use, copy, modify, and/or distribute this software for any purpose with or without fee is hereby granted, provided that the above copyright notice and this permission notice appear in all copies. The software is provided "as is" without warranty of any kind.
Respiro-en (breath remover)
The breath remover finds breaths with Respiro-en by Dong Yang, Tomoki Koriyama and Yuki Saito ("Frame-Wise Breath Detection with Self-Training", INTERSPEECH 2024, arXiv 2402.00288), code and weights under the MIT License, Copyright (c) 2024 DongYANG (github.com/ydqmkkx/Respiro-en); the MIT permission notice above applies to it. The model was trained on LibriTTS-R by Yuma Koizumi et al. (openslr.org/141), Creative Commons Attribution 4.0 (creativecommons.org/licenses/by/4.0). We exported its weights to ONNX and quantised them to 8 bits; it runs in your browser and nothing is uploaded.
RNNoise (noise and clean tools)
The clean tools' on-device speech cleaner, used when our server cannot clean a file, is RNNoise by Jean-Marc Valin and Xiph.Org, BSD 3-Clause License (github.com/xiph/rnnoise). Copyright (c) 2017, Mozilla; 2007-2017, Jean-Marc Valin; 2005-2017, Xiph.Org Foundation; 2003-2004, Mark Borgerding. Redistribution and use in source and binary forms, with or without modification, are permitted provided that the copyright notice, this list of conditions and the disclaimer are retained. The WebAssembly build and its loader come from @jitsi/rnnoise-wasm 0.2.1, Apache License 2.0. It runs in your browser; nothing is uploaded.
3. LGPL components
Rifflo delivers one LGPL-licensed component to your browser, and only when you export an MP3.
LAME MP3 encoder (inside @mediabunny/mp3-encoder)
MP3 export uses the LAME 3.100 encoder, licensed under the GNU Lesser General Public License, version 2.1 or later (lame.sourceforge.io/license.txt, licence text at gnu.org/licenses/old-licenses/lgpl-2.1). LAME's source is at lame.sourceforge.io and the WebAssembly build we use, with its build instructions, at github.com/Vanilagy/mediabunny/tree/main/packages/mp3-encoder. We use it without modification. The LAME project asks to be credited with a link; this is that credit.
How the LGPL's relinking requirement is met: the LAME build is not compiled into Rifflo's own code. It is served as one separate file, mediabunny-mp3-encoder-<hash>.js under /assets/, which Rifflo fetches and evaluates the first time you export an MP3. To substitute your own build of LAME, replace that one file with another build that exposes the same MediabunnyMp3Encoder.registerMp3Encoder() entry point; nothing else in Rifflo changes. The encoder runs entirely in your browser and no audio is sent to us.
The bridge exposes constant-bitrate encoding (128, 192, 256 and 320 kbit/s); variable-bitrate presets are not available in this build.
No GPL-licensed component is delivered. AAC and Opus export use the encoder built into your browser; we do not ship one.
4. Fonts and icons
| Asset | Licence | Source |
|---|---|---|
| Inter (variable), via @fontsource-variable/inter 5.3.0 | SIL Open Font License 1.1 | rsms.me/inter |
Source Serif 4 (variable), via @fontsource-variable/source-serif-4 | SIL Open Font License 1.1 | github.com/adobe-fonts/source-serif |
| Libre Baskerville, via @fontsource/libre-baskerville | SIL Open Font License 1.1 | github.com/impallari/Libre-Baskerville |
| Lucide icons, via lucide-react 0.511.0 | ISC | lucide.dev |
Platform brand marks (Spotify, Apple Music, YouTube, TIDAL, Amazon Music, Deezer, SoundCloud), seven SVG paths copied from Simple Icons 16.36.0 | CC0 1.0 for the icon data; the marks themselves remain trademarks of their owners | github.com/simple-icons/simple-icons |
The SIL Open Font License text is at openfontlicense.org. Fonts are served from our own domain, not from a third-party font service.
The platform marks appear small and in a single colour, only to identify which streaming service a loudness number or a pasted link refers to. Rifflo is not affiliated with or endorsed by any of these services; see the Simple Icons disclaimer.
5. Application template and other dependencies
Rifflo is built on the ShipAny TanStack template, which is proprietary software used under a commercial licence from its authors. It is not open source, its copyright notices are retained in our source code, and the Rifflo source is not available for redistribution.
The application also depends on widely used open-source libraries that run in your browser or on our servers, including React, TanStack Router, Start, Query and Form, Tailwind CSS, shadcn/ui, Drizzle ORM, better-auth and Paraglide JS. They are licensed under permissive licences (MIT, ISC or Apache 2.0) that require preservation of their copyright notices, which we keep in the distributed code. A complete machine-readable list of dependencies and licences for the current build is available on request from support@voice-remover.com.
6. Demo track
The demo you can play on the Rifflo home page is an excerpt of:
"Not Alone" by Fredy V, mix A by students of the McGill Graduate Program in Sound Recording, obtained through the Open Multitrack Testbed (multitrack.eecs.qmul.ac.uk). Licensed under Creative Commons Attribution 3.0 (CC BY 3.0).
Changes: the excerpt has been shortened and processed through Rifflo's mastering chain for demonstration; the "before" version is the unaltered mix. The licensor does not endorse Rifflo.
If you use the Open Multitrack Testbed, please cite: B. De Man, B. Leonard, R. King, J. D. Reiss, "An analysis and evaluation of audio features for multitrack music mixtures", 15th International Society for Music Information Retrieval Conference (ISMIR 2014).
No other third-party recordings are used in the public product. Material used only in internal tests is not distributed.
7. Models run in the cloud
These tools send your file or text to a processing partner and run a model there; nothing from these models is delivered to your browser. We list them because their licences govern what you may do with the results.
| Tool | Model | Licence | What it means for you |
|---|---|---|---|
Stem splitter, vocal remover, karaoke, acapella, instrumental, background-music, drum, bass, piano and guitar remover; the instrumental of the lead and backing splitter; stem-aware instructions in the studio | BS-RoFormer, hosted as cottom/bs-roformer , a personal model on Replicate | Unknown. The host publishes no source and states no licence for the code or the weights. | You keep every right you had in the recording you uploaded; we make no claim on the stems. We cannot pass on a licence for the model because none is published. |
| The same separations, whenever that model fails | Demucs v4 (htdemucs_ft, htdemucs_6s), hosted as cjwbw/demucs | Code MIT ( github.com/facebookresearch/demucs ). The model weights are released by their author "only for scientific purposes"; the six-stem weights are described as likely CC BY-NC. | The separation itself is research-licensed. You keep every right you had in the recording you uploaded; we make no claim on the stems. If you need a written commercial licence for the model, we do not have one to pass on. |
| Lead and backing vocal splitter (the two vocal files) | KIE stem split, ai-music-api/separate-vocals . KIE does not name the model. | Closed service. KIE's terms do not address who owns the output. | We make no claim on the files. We cannot pass on a licence for the model because none is published. The instrumental on this page comes from the separation models above. |
Dialogue remover, speech, music and effects splitter, remove background music from video, the speech modes of the audio separator and Pro separation (as a guide; the stems above do the rest) | SAM-Audio (Meta), large model, hosted as geopti/sam-audio-large | SAM License (Meta): use, copy and modify allowed, including commercially; no reverse engineering; research publications must credit it. | We make no claim on the files. Its output is only used to decide, for each moment and pitch, how much of your own audio belongs to the speech or the music; the files you download are cut from your recording, not generated. |
Violin extractor, string instrument remover, wind instrument remover | MVSep Mega 53 stems (BS-RoFormer) by ZFTurbo, release v1.0.21 of Music-Source-Separation-Training , run by us as a private model on Replicate | Code MIT (Copyright (c) 2024 Roman Solovyev (ZFTurbo)). The weights are published in the same MIT-licensed repository's releases with no separate licence stated for them. | You keep every right you had in the recording you uploaded; we make no claim on the files. The second file is your recording minus the first, so the two add back up to it. We cannot pass on a licence for the weights beyond what the repository states. |
Noise remover (speech), the speech in the white, static, background, wind and crowd noise removers, the audio cleaner and the podcast enhancer, and the first stage of the speech and audio enhancers | NVIDIA NemotronLabs RE-USE ( nvidia/NVIDIA-NemotronLabs-RE-USE ), run by us as a private model on Replicate | OpenMDW License Agreement, version 1.1 (model, code and documentation). Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved. The licence text ships with our copy of the model. | The licence puts no restriction on the outputs. |
| The same speech cleanup, only when the model above fails | MossFormer2 SE 48 kHz (ClearerVoice-Studio), hosted as playmore/speech-enhancer | Apache License 2.0 (code and weights) | No restriction on the output. |
| Speech enhancer, audio enhancer (after the speech cleanup above) | Resemble Enhance, hosted as resemble-ai/resemble-enhance | MIT (code and weights; github.com/resemble-ai/resemble-enhance ) | No restriction on the output. |
| Text to audio samples | Stable Audio Open 1.0, hosted as stackadoc/stable-audio-open-1.0 | Stability AI Community License . The page carries the required "Powered by Stability AI" line. | You own the samples you generate. The licence is free for organisations under USD 1 million in annual revenue; above that, Stability AI requires a commercial licence from them. |
Music noise removal, the echo remover and the reverb remover run in your browser with Rifflo's own code and send nothing anywhere. Replicate deletes uploads after 24 hours and outputs after 1 hour; KIE deletes uploads after 24 hours and keeps results for 14 days; we keep a copy of the results for 24 hours, keyed by the file's checksum, so a repeat of the same file in that time is not processed twice.
8. Questions or corrections
If you believe a component is missing from this page or a licence is misstated, write to support@voice-remover.com. We correct this page promptly.