AI music artifact remover. Suno stems, without the metallic ring.

Because the neural codec inside AI music generators leaves structured residue in the decoded audio: metallic spectral peaks locked to a roughly 200 Hz grid, plus a flat hiss floor just under the codec cutoff. That is what you hear as metallic ring, watery shimmer, splashy cymbals and a brittle top end. Honenuki finds that grid and notches only the detected peaks, so the ring goes and the music stays.

Honenuki looks for two independent codec fingerprints: metallic spectral peaks locked to a roughly 200 Hz grid, and a flat static noise floor just below the codec cutoff. It reports a verdict for every upload, and the verdict is display-only — it never blocks or forces processing.

Suno, Udio, Mureka and other generators built on the same family of neural codecs. Honenuki keys on the codec fingerprint rather than on a brand, so tracks from comparable generators clean up the same way. Full mixes and exported stems both work.

Either works — Honenuki cleans the file you give it. It is not a stem separator: it does not split a song into vocals, drums and bass. If you already have stems, cleaning them one by one gives the best result, because vocal stems usually carry the strongest artifacts.

No. Honenuki attenuates only narrow bands around detected peaks instead of pulling the whole high end down, and a restoration stage — saturation, de-esser, air — brings texture and brightness back. Drum stems use a lower high-frequency cap with less attenuation around transients, so attacks and cymbals survive.

Yes. An upload with no codec fingerprint is labelled likely human-made and left as it is; the verdict informs and never blocks. You can compare original, processed and removed-residual at the same playback position at any time.

Before. Codec artifacts and the high-frequency hiss floor get louder along with everything else when you raise the level, so clean first and master afterwards. Honenuki is not a mastering chain and applies no limiting or loudness normalisation.

WAV, MP3, FLAC and AIFF. Cleaned stems are rendered as WAV and delivered together in a single ZIP; MP3 export is also available. What you hear in the browser is the same server render you download — there is no preview-only processing.

No. Honenuki removes codec artifacts you can hear. It does not strip inaudible watermarks such as SynthID or C2PA credentials, and we make no claim about distributor or streaming-platform screening. Its own verdict is shown for information only.

Not in the prompt sense. Excluding a genre, instrument or vocal type happens in Suno's own Exclude field, before the song is generated. Honenuki works on the exported audio: it removes the codec artifacts baked into it and leaves the arrangement, melody and lyrics exactly as generated.

The web app is free during early access ($0; regular price $6.99 per month, 6,900 KRW in Korea). A one-time VST licence at $119 / 119,000 KRW is in preparation and not released yet. Any pricing change is announced before it takes effect.