Free audio tag reader · nothing is uploaded
What is this audio file declaring?
AI music and AI voices have a metadata problem of their own: the disclosure lives — or doesn't — in ID3 tags and WAV chunks almost nobody inspects. Drop an MP3 or WAV and see exactly what its tags declare, byte-level, in your browser.
Check an audio file
or try a sample:
Why audio provenance is its own problem
The image world got C2PA, platform labels, and a public argument about all of it. Audio got the same rules — the EU's machine-readable marking duty and California's latent-disclosure duty both name audio explicitly — and a fraction of the tooling. What an MP3 declares rides in ID3 frames designed for song titles; what a WAV declares rides in INFO fields designed for studio notes. Generators use them inconsistently: some stamp a user-defined frame with their name, some write an encoder string, most write nothing at all. This checker reads what is actually there — every frame and field, decoded — and applies the same disclosure-copy rules as the image tools, so a comment that says the track was created with AI tools counts as what it is: a disclosure.
It also strips. If you are publishing audio and want a clean file — or you received one and want to see what its tags were carrying before you decide — the strip download removes the tag structures and nothing else: the audio payload is preserved byte for byte, and the tests behind this page prove it on every run.
Frequently asked questions
Where does AI disclosure live in an audio file?
In the metadata tags, in the same spirit as images: an MP3's ID3 frames (a user-defined TXXX frame naming the generator, a comment stating the track was AI-created, encoder fields), or a WAV's INFO comment and broadcast-extension (bext) description. Some generators write these automatically; many write nothing, which is itself the finding this tool reports.
If no disclosure is found, is the audio human-made?
No — and this tool says so in its own output. Absent tags prove only absent declarations. A fully synthetic track can carry no tags at all, and a human recording can carry a comment mentioning AI tools used in mixing. Metadata reading answers the provenance question 'what does this file declare?', never the forensic question 'how was this really made?'
Does stripping the tags damage the audio?
No. The stripper removes the tag structures at the byte level — the whole ID3v2 tag and ID3v1 trailer on an MP3, the INFO and bext chunks on a WAV (with the RIFF size field corrected) — and copies the audio frames and sample data untouched. The tool's own tests prove the stripped file's audio payload is byte-identical to the original's.
What is an AIGC marker?
AIGC — AI-generated content — is the vocabulary of China's labeling regime, in force since 1 September 2025, which requires explicit and implicit labels, including content identification in file metadata. A tag field carrying an AIGC marker is a disclosure signal this checker recognizes and reports.