Blush Measures Suno Decoder Imprint In Generated Tracks
A new tool called Blush detects a high-frequency pattern left by Suno's decoder. The author found the same faint imprint in 36 of 40 Suno tracks and none of 42 human recordings or 53 Udio tracks from 2024.
1 source 3h ago
Reporting from 1 source: GameBusiness.jp.
Why it matters
Blush points at the decoder artifact as the real detection target and reports mixed and processed tracks section by section, the two weak points the cited detection efforts say they have not solved.
Blush measures the mark in time and frequency bands, follows it through mixing and processing, and reports which segments came from which source.
The author built Blush after nearly three years of making songs with Suno and working on tools that correct human vocals. At that point the boundary between AI output and human performance inside a single file stops being obvious, even to the person who made it. The 60-millisecond high-band mark is described as a trait of the decoder family that assembles audio in 48kHz steps of 60 milliseconds, not a watermark Suno planted. The same shape appeared in another generator, YuE2. Blush splits a track into seven layers including high-band cutoff, noise floor, envelope modulation, stereo correlation and waveform repetition, and shows the numbers and figures behind each call.
Written with AI by Yomimono from the 1 cited source below, including Japanese-language reporting where cited. See our editorial policy.
Add Yomimono as a preferred source on Google
Sources
- GameBusiness.jp Sunoが使われた曲を見分けるツール「Blush」を作った（CloseBox）