Neiro
What it is
Neiro is a local audio worksuite for separating sources, repairing recordings, transcribing audio and editing waveforms. Audio processing stays on the user's machine. It offers a Tauri desktop app and a browser interface launched with `neiro ui`. Separation options include vocals, instrumentals, harmonic and percussive parts, four- or six-stem mixes, and drum kits; each result includes a null-test residual. Restoration tools address clipping, hum, noise, reverberation and bandwidth extension, as well as reference mastering. Transcription can export MIDI, MusicXML, ASCII tablature and LRC lyrics. Studio supports non-destructive waveform edits, while Learn includes loop regions, count-in, metronome, step mode, WebMIDI and DAW wait mode. VST2 and CLAP injectors can bring DAW audio into Neiro. Core DSP works without model downloads, while optional neural backends download weights on first use. The interface binds to 127.0.0.1, and the Python package requires Python 3.10–3.12. Neiro is free; model licenses vary, and some restrict use to non-commercial or research settings.
For and against
For
- Audio processing stays on the user's machine.
- Core DSP works without model downloads.
- Restoration includes declipping, hum removal and denoising.
- Exports transcription to MIDI and other formats.
- Supports non-destructive edits and DAW audio capture.
Against
- Neural weights download on first use.
- Some model licenses restrict use to non-commercial or research settings.
- Compressed or video inputs require ffmpeg on PATH.
- Granted adapter plugins run without a sandbox.
The verdict
Neiro brings separation, restoration, transcription and editing into a local audio workflow. Check individual model licenses and the Python and ffmpeg requirements before relying on particular inputs or backends.
Who it suits
Neiro suits people who want local audio separation, restoration, transcription or waveform editing. Its practice tools and DAW capture options may also suit musicians working with audio on a desktop.
On the record
- DAW integration
- Shared-window VST2 and CLAP injectors can capture audio into Neiro's interface.github.com · read 29 Sep 2026
- Desktop downloads
- The release page lists Windows MSI/EXE, macOS DMG, and Linux AppImage/DEB installers.github.com · read 30 Sep 2026
- Editing and practice
- Its Studio supports non-destructive audio edits, while Learn offers loop regions, count-in, metronome, WebMIDI, and DAW wait mode.github.com · read 30 Sep 2026
- Evaluation limits
- Full MUSDB18-HQ and MAESTRO evaluation numbers require user-provisioned datasets.github.com · read 30 Sep 2026
- Extension limits
- Neiro documents a local Python adapter plugin MVP, but granted adapters run in the Neiro process without a sandbox.github.com · read 30 Sep 2026
- Integrations
- The project documents VST2 and CLAP injectors for shared-window DAW capture, and a VST2 effect that works as a pass-through injector in a DAW.github.com · read 30 Sep 2026
- Interfaces
- The worksuite is available as a Tauri desktop app or through `neiro ui` in a browser, using the same modules and local cache.github.com · read 30 Sep 2026
- License
- The repository license is MIT for the Neiro engine source, while model weights have separate licenses.github.com · read 30 Sep 2026
- Local interface security
- The UI binds to 127.0.0.1, and the desktop shell restricts its connections to the local engine origin.github.com · read 30 Sep 2026
- Local network boundary
- The interface binds to 127.0.0.1, and the security policy says the app has no outbound network activity by default apart from user-initiated model downloads and updates.github.com · read 29 Sep 2026
- Local processing
- The maker says audio is processed locally and not transmitted; outbound network activity is limited to user-initiated model downloads and application updates.github.com · read 30 Sep 2026
- Model downloads and licensing
- Neural weights are not bundled with desktop releases and download on first use; each model carries its own license, which can include non-commercial or research-only terms.github.com · read 30 Sep 2026
- Model options
- The core DSP floor works without model downloads; neural backends such as Demucs, Basic Pitch, and AudioSR are optional.github.com · read 29 Sep 2026
- Model security
- Model weight downloads are checked against manifest SHA-256 values, and the security policy warns that third-party weights can be dangerous.github.com · read 29 Sep 2026
- Purpose
- Neiro is a local audio worksuite for source separation, restoration, transcription, and audio editing.github.com · read 30 Sep 2026
- Requirements
- The Python package requires Python 3.10–3.12, and compressed or video inputs require ffmpeg on PATH; WAV and FLAC work without it.github.com · read 29 Sep 2026
- Restoration
- Its restoration features include declipping, hum removal, denoising, dereverberation, bandwidth extension, and reference mastering.github.com · read 30 Sep 2026
- Separation
- It separates vocals, karaoke, harmonic and percussive parts, and 4- or 6-stem audio, and provides a null-test residual.github.com · read 30 Sep 2026
