Neiro

Overview

Neiro is a local audio suite for musicians and anyone repairing or working with recordings. It is best suited to users who want restoration, source separation, transcription, editing, and practice tools in one workflow. Its breadth and on-device processing are compelling, though optional model licenses and downloads deserve attention before use.

Key features

Audio repair and separation

Neiro covers declipping, hum removal, denoising, dereverberation, bandwidth extension, and reference mastering, alongside noise reduction, click and crackle removal, and batch processing. That range makes it useful for varied cleanup jobs without limiting the app to restoration. It can also split audio into vocals, instrumentals, harmonic and percussive parts, four- or six-stem mixes, and drum kits. A null-test residual accompanies each separation result, giving users a way to inspect what remains rather than treating a split as a black box.

Transcription, editing, and practice

Transcription to MIDI, with MusicXML, ASCII tablature, and LRC lyrics exports, gives Neiro a role beyond repair. Studio's non-destructive waveform edits suit users who want to make changes without replacing the original audio. Learn adds loop regions, count-in, metronome, step mode, WebMIDI, and DAW wait mode for practice workflows. The combination is unusually broad, but users seeking only a focused repair utility may not need the wider toolkit.

DAW capture and models

VST2 and CLAP injectors can capture audio from a shared DAW window into Neiro, and a documented VST2 effect acts as a pass-through injector. The core DSP floor works without model downloads; neural backends such as Demucs, Basic Pitch, and AudioSR are optional and their weights download on first use. The project checks downloaded weights against manifest SHA-256 values, but warns that third-party weights can be dangerous. Each model retains its own license, which may be non-commercial or research-only, so users should check terms before adopting a model in commercial work.

Pricing

Neiro is free, with a free plan and no paid tier described. Its core DSP functions work without model downloads, while neural weights are not bundled with desktop releases and download when first used. That keeps the base workflow accessible, but users should account for model-specific licensing and download requirements if they need neural processing.

Platforms

Neiro supports Linux, macOS, Windows, self-hosted use, and a browser interface launched with neiro ui. Desktop releases include Windows MSI/EXE, macOS DMG, and Linux AppImage/DEB installers. The UI binds to 127.0.0.1, and the desktop shell restricts connections to the local engine origin. Audio stays on the user's machine; the security policy says there is no outbound network activity by default apart from user-initiated model downloads and updates.

The Python package requires Python 3.10–3.12. Compressed or video inputs require ffmpeg on PATH, while WAV and FLAC work without it. This is a practical constraint for users who want to process those formats through the Python package.

Who it's for

Neiro suits musicians and audio editors who want local processing across repair, separation, transcription, waveform editing, or music practice, particularly those willing to manage optional model downloads. It is less suited to users who need a single-purpose repair app without model-license checks, or those who require full evaluation benchmarks without supplying datasets: full MUSDB18-HQ and MAESTRO evaluation numbers require user-provisioned datasets.

Support comes through documentation and public GitHub Discussions or Issues, with private reporting for security vulnerabilities. The engine, desktop shell, and frontend are MIT licensed, but that does not override the separate licenses attached to individual models. Neiro documents a local Python adapter plugin MVP; granted adapters run inside the Neiro process without a sandbox, so that extension path calls for care.

Pros and cons

• Pros: Broad restoration and separation tools, plus transcription, non-destructive editing, and practice features, give users multiple audio workflows in one suite.
• Pros: Local processing and a localhost-bound interface keep the normal workflow on the user's machine.
• Pros: Core DSP works without model downloads, so neural weights are not a prerequisite for the basic processing floor.
• Cons: Optional neural weights have individual licenses, some non-commercial or research-only, and download on first use.
• Cons: Compressed or video input through the Python package requires ffmpeg, unlike WAV and FLAC.
• Cons: Python adapters run without a sandbox, and full MUSDB18-HQ and MAESTRO evaluation numbers depend on user-provisioned datasets.

Alternatives

For a focused restoration comparison, browse Audio Restoration Software.

• Cathar is a free option for Linux and macOS users.
• CEDAR Cambridge is a paid modular hardware and software system for macOS and Windows, with custom pricing and configurations that vary.
• VinylRest is a free choice for web, macOS, and Linux.
• Wave Corrector is a free option for Windows and Linux.
• PD Cleaner is a free Windows option.
• Wave Arts Master Restoration Suite 6 costs 99.00 USD per once and offers five restoration plug-ins; choose it if a host-based plug-in suite is a better fit. It requires a host that supports audio plug-ins.
• SpectraLayers Pro is a paid macOS and Windows alternative with spectral audio editing, audio repair and restoration, and AI-assisted processing.
• LANDR ReHance is a paid macOS, web, and Windows choice for users who want ReHance through LANDR Studio; the annual-billed plan is 11.99 USD per month and requires a LANDR Studio subscription.

Verdict

Choose Neiro if you want a free, local audio workspace that spans repair, separation, transcription, editing, and practice. Its strongest case is that range without requiring neural models for core DSP. Look elsewhere if your workflow depends on a narrowly focused restoration tool, or if model licensing, downloads, or unsandboxed adapters are unacceptable constraints.