Upload the song
Use the highest-quality file you have. A WAV or 320 kbps MP3 gives the clearest read.
Check whether a song was sung, played or fully generated by AI. The report scores the full mix, the vocal and the instrumental separately, and names the generator it most resembles.
Use the highest-quality file you have. A WAV or 320 kbps MP3 gives the clearest read.
Full mix, vocals and instrumental. A gap between them usually means AI was used for one part only.
Use the verdict band to decide whether to accept, ask the artist, or pass — not as an automatic rejection.
Different teams, same question: was this recording made by a person?
Check demos before signing or advancing, and keep the report with the deal file.
Triage submissions quickly and ask artists about tracks that land in the AI bands.
Confirm what you are licensing, especially when a brief excludes AI-generated music.
Enforce "human-made only" rules consistently across hundreds of entries.
Check coursework submissions for fully generated tracks.
Curious whether that new artist in your feed is real? Check a song for free.
Many songs now mix human and AI parts: a real singer over a generated beat, or an AI voice on a produced track. A single score hides that. Separate stem reads show which part carries the AI signal.
The audio carries strong generator fingerprints across the track.
Likely source
Timeline
Stems
Detection is probabilistic. Use it to route tracks to review, not as the sole basis for takedowns. How to read results
When the engine returns separate stems, the report shows a vocal read and an instrumental read next to the full-mix score. A human instrumental with AI vocals — or the reverse — shows up as a gap between the two.
Related
Creator gives you 200 checks a month with history in your dashboard.