Remove tags while preserving the original sound
Audio files can carry artist names, comments, generator information and portable provenance alongside their sound. Free AI Wiper can inspect supported WAV, MP3, native FLAC and M4A metadata and remove selected records without converting the file or re-encoding its audio.
The original audio bytes are preserved. Removing a tag therefore does not change the recording's quality, bitrate or duration. The audio format page lists the supported tags and layouts.
How to clean audio metadata
- Keep a copy of your original file, then select Audio in the cleaner.
- Choose a WAV, MP3, native FLAC or supported M4A file of up to 50 MiB.
- Review the findings. Supported AI provenance is shown separately from ordinary descriptive metadata.
- Select private metadata removal if you also want supported titles, artist information, comments or pictures removed. Otherwise those records stay.
- Clean the file, read the after-report and download the verified output. Check playback in a compatible player.
Processing happens locally. A failed browser preview does not prevent metadata processing. An unsupported structure or failed verification produces an error rather than a clean result.
Clean M4A tags without converting to MP3
For supported audio-only M4A files, the cleaner clears recognized C2PA/XMP carriers and movie-level tags such as AIGC labels, generation parameters and supported software fields. Titles and other ordinary tags remain unless private metadata removal is selected. The underlying AAC or ALAC stream stays byte-identical, and file offsets do not move.
Support requires an M4A major brand and self-contained AAC/ALAC tracks. Encrypted, fragmented, externally referenced and mixed video/audio files are refused. Other brands, codecs and metadata locations are outside this supported path. See the audio format boundaries before interpreting an empty findings list.
Why an AI-generated WAV may have no metadata findings
Our supplied Gemini WAV regression sample is 295,724 bytes. Its RIFF container contains only a 16-byte audio-format chunk and a 295,680-byte audio-data chunk. The cleaner reports no supported metadata findings and returns the file byte-for-byte unchanged.
That result describes its container metadata. It cannot establish whether the recording contains an invisible audio watermark. Google's SynthID documentation describes watermarks embedded in generated audio; the Audio tab does not detect or remove those signals.
Do not treat a metadata-clean report as proof that sound is human-made or that another detector will classify it differently. See AI metadata versus watermarks for the distinction.
Content Credentials and AIGC labels
For WAV, the cleaner recognizes the C2PA RIFF carrier and AIGC chunk. For MP3, it recognizes supported C2PA GEOB frames and AIGC user-text frames. For FLAC, it handles AIGC Vorbis comments and supported leading ID3 C2PA carriers.
These are detected records, not validated signatures. The C2PA specification defines ID3 and RIFF manifest carriers. Unknown metadata and external provenance associations may remain after supported records are removed. Content Credentials verification is a separate task.
Why file size may stay the same
WAV chunks are replaced with empty padding to preserve offsets. MP3 ID3 frames are rebuilt within the original tag space, with unused space cleared. FLAC comments can shrink while the encoded frames stay unchanged. Selected M4A records are cleared in place and become padding, so a cleaned M4A usually keeps its original size. File size alone is not a reliable measure of removal.
The cleaner inspects the output again and verifies that changes match the selected metadata edits and that protected audio bytes remain identical. Our test method explains the fixture checks and their limits.