Independent audio provenance analysis
Hear the signal behind the song.
Upload a recording and find out whether its vocals or instrumentation show patterns typical of AI-generated music.
THE OUTPUT
A classification, separate vocal and instrumental readings, a confidence rating, and per-window evidence — never a claim of proof.
Start analysisLive analysis console / 01
Analyze a track
Drop zone / encrypted transfer
Drop a song here
Or select a file from this device. It will be sent only when you run the check.
- MP3
- WAV
- FLAC
- AAC
- M4A
- MP4
- OGG
- OPUS
MP3, WAV, FLAC, AAC, M4A, MP4, OGG, OPUS · max 25 MB maximum · 5 seconds to 15 minutes
Running the check sends the complete file over HTTPS to our server and classification provider. We do not save the audio or publish the result. Submit only material you are permitted to process; the report grants no rights in the recording.
- No cost
- Results in seconds
- Encrypted upload
- No sign-up
- Files
- MP3, WAV, FLAC, AAC, M4A, MP4, OGG or OPUS
- Limits
- 25 MB · 5 seconds to 15 minutes
- Access
- Free · no account · 12 checks per rolling 24 hours
- Privacy
- External classification · no audio saved by this site
Process / 02
From file to report
The homepage keeps the process brief; the full request lifecycle is documented separately.
1. Choose
Select a supported file. Your browser checks that it can be read and measures its duration.
2. Send
When you run the check, the complete file travels over HTTPS to our server.
3. Classify
A specialist provider assesses the recording and returns its category and supporting fields.
4. Review
Read the category with the component indicators, confidence and available timeline—not as proof on its own.
Generator context / 03
Generator context, not attribution
Different generators can leave different characteristics, and their output changes as models evolve. You can analyse exports from platforms such as Suno, Udio or ElevenLabs Music, but this detector does not identify the platform or model that made a track.
Interpretation / 04
Understanding your result
Category
The provider's primary result: AI vocals, AI instrumental, not AI, or inconclusive.
Supporting indicators
Vocal and instrumental percentages remain separate; this site does not blend them into one score.
Confidence
Low, Medium or High describes support for the reading, not measured accuracy or certainty.
Timeline values may show where the assessment changes across a recording. None of these fields establishes authorship, intent or the production process.
Read the result guidePractical use / 05
Why check a track?
Use the report to decide what to review next—not to automate a conclusion.
- Compare a collaboration with its disclosure and project records.
- Screen a submission before requesting source files or more context.
- Preserve one technical observation during a provenance dispute.
- Document an editorial or research question with the exact file and date.
Responsible interpretation / 06
What the analysis cannot settle
- Audio classification has uncertainty; this site has not published a controlled accuracy percentage for the current system.
- Compression, re-encoding, mastering, source quality and short or unrepresentative excerpts can change the available evidence.
- Hybrid workflows can place performed, edited and generated material in the same finished mix.
- A result cannot prove who made a track, why it was made, which tool was used, or whether a policy was broken.
Access and data / 07
Free to use, with a clear data boundary
Access
There is no charge, registration, email gate or paid result tier. Everyone uses the same classification path, with a fair-use limit of 12 analyses per rolling 24 hours.
Privacy
The complete file goes to our server and classification provider. We do not save the audio. Numeric results can be reused for 14 days, not a verified deletion deadline; duration and a network-address fingerprint are logged. Read the privacy policy for records and provider safeguards.
There is no public report link or personal result history. A PDF is created in your browser only if you choose to download it.
Read the privacy policyQuestions / 08
Frequently asked questions
The complete audio file is sent to a specialist classification service, which returns a category, separate vocal and instrumental indicators, confidence and, when available, timeline values. These are patterns in the submitted recording, not a reconstruction of how it was made.
Learn more / 09
Guides and research
Ready when you are