Open-source licences
Last updated: 4 September 2026
1. Noise Filter components
DeepFilterNet — denoising
Separates the voice from everything else the mic picks up. DeepFilterNet, copyright © Hendrik Schröter and contributors. Distributed under your choice of MIT or Apache-2.0; the model weights ship under the same licence and were trained on the DNS-Challenge corpus.
Silero VAD — speech detection
Decides whether speech is present, working on the already-cleaned signal — this is what guarantees total silence when you play without talking. Silero VAD, copyright © Silero Team, MIT licence.
tract — inference engine
Both models run on tract (Sonos), an ONNX engine written in Rust, under MIT or Apache-2.0. No extra native dependency is installed on your machine.
2. Yes, this is artificial intelligence
These are pre-trained neural networks. Three things worth stating plainly:
- they run on your machine, never in the cloud;
- your voice trains nobody — the models are frozen, shipped as they are;
- they are free and will stay so: no third-party subscription is involved.
3. Licence texts
The full licence texts ship with the software, as those licences require: open the Audio Engine window and click “Licences”. The Apache-2.0 text is also available on apache.org.
4. Contact
Any question about these components or their use: contact@jamodio.com.