NaSuno is an MIT-licensed Python and C++ audio DSP toolkit for musicians, producers, and sound designers that removes synthetic watermarks and forensic provenance markers from AI-generated audio while aiming to leave the audible result intact.

What it is

NaSuno is a defensive signal-processing framework that lives in the audio-processing and DSP ecosystem. It is built on ISO 226 equal-loudness contours, Bark-scale critical-band psychoacoustics, and zero-phase IIR notch filtering, and it is designed to decouple human musical creativity from synthetic platform tracking. The toolkit is organised as a public wrapper named NaSuno, which hosts the browser-based Web Audio DSP studio, mathematical research papers, forensic specifications, and architecture rules, alongside a private engine named NaSuno-core that contains the high-throughput C++ and Python DSP pipeline and an automated batch-processing CLI.

The concrete problem it addresses is the marker layer that generative audio systems leave behind. Suno, Udio, ElevenLabs, and Google SynthID inject proprietary synthetic watermarks, high-frequency ultrasonic carriers at 19.5 kHz, phase-locked spread spectrums, and latent provenance tags into generated waveforms. When a producer folds an AI stem, sound effect, or background texture into a human composition, the entire track inherits those forensic tags. NaSuno targets those carriers and the container metadata that ships them, including C2PA provenance manifests, RIFF INFO chunks, ID3 tags, and BWF acoustic metadata, using bi-directional forward-backward IIR filtering with Q values of 30 or higher rather than blunt removal, so that transients and upper harmonics are preserved.

Key capabilities

• Ultrasonic carrier notch filtering at 19.5 kHz using zero-phase bi-directional IIR filters, dropping the tracking carrier below the -110 dBFS noise floor while isolating cymbal air and vocal breath.
• An ISO 226 perceptual masking engine that evaluates instantaneous masking thresholds across all 24 Bark-scale critical bands so anti-attribution dither and micro-dispersion stay inside acoustically invisible corridors.
• Latent and sub-band dispersion, including spread-spectrum neutralization that scrambles pseudo-random watermark chip sequences across mid-range acoustic formants.
• Phase de-correlation of multi-band phase anomalies while preserving mono compatibility and stereo field width.
• A deep metadata scrubber that strips C2PA manifests, RIFF INFO chunks, ID3 tags, and BWF acoustic metadata from the container wrapper.
• TPDF dither reconstruction with output to WAV, FLAC, MP3, and OGG.
• A published 14-vector anti-attribution matrix that records each target artifact, its forensic carrier mechanism, the corresponding DSP countermeasure, and the reported evasion status.

Who uses it and how

• Producers and sound designers who incorporate AI stems, sound effects, or background textures into human compositions and want the finished track free of forensic tags before distribution.
• Musicians working with Suno, Udio, and ElevenLabs output who need proprietary watermark carriers and latent provenance tags neutralized without audible artefacts.
• Mixing and mastering workflows that require the stereo field width and mono compatibility of the source to survive processing.
• Music production and FL Studio-adjacent workflows that exchange rendered files in WAV, FLAC, MP3, or OGG form.
• Engineering and research readers who want the forensic specifications, research papers, and architecture rules that the public wrapper publishes.

Getting started

The documented entry point is the hosted Web Audio Studio at the project's GitHub Pages site, https://ahmadhassan-bted.github.io/NaSuno/, which launches the DSP studio in the browser. The README excerpt provided here does not name a package, Docker image, or compose file for a local installation.

How it compares

The registry facts name no directly comparable watermark-removal or anti-attribution tool; the systems listed, Suno, Udio, ElevenLabs, and Google SynthID, are the watermark injectors NaSuno is built to counter rather than alternatives to it. On the evidence available, it stands alone in this registry.

When to use it — and when not to

The public repository is a wrapper around a hosted studio and written specifications, while the high-throughput DSP pipeline and batch-processing CLI sit in the private NaSuno-core engine, so anyone who needs to inspect, audit, or self-host the full processing engine will not find it in the public material. The README excerpt is also truncated and supplies no dependency list, installation command, or usage walkthrough, which makes it a poor starting point for a team that needs a documented library to embed. It best suits users who want the browser studio and the published forensic methodology, not a fully open, reproducible processing stack.

┌────────────────────────────────────────────────────────────────────────────────────────────────────────┐
│                                       NASUNO CORE CAPABILITIES                                         │
├───────────────────────────────┬───────────────────────────────┬───────────────────────────────┬────────┤
│ 🚨 Watermark & Carrier Purge  │   Psychoacoustic Masking      │   Forensic Anti-Attribution   │    WAV │
│  - Ultrasonic Notch (19.5kHz) │  - ISO 226 Equal Loudness     │  - Suno AI Carrier Excision   │  - FLAC│
│  - SynthID Inaudible Spread   │  - Bark-Scale Masking Latency │  - Udio Latent Neutralization │  - MP3 │
│  - Zero-Phase Bi-Directional  │  - TPDF Dither Reconstruction │  - C2PA / RIFF Tag Stripping  │  - OGG │
└───────────────────────────────┴───────────────────────────────┴───────────────────────────────┴────────┘