THE DETAILS BEHIND THE CLEAN COPY
Less metadata.
Clear expectations.
Quiet Media removes supported descriptive metadata. It preserves the information a file needs to display or play correctly.
Images: a fresh PNG
JPG, PNG, and WebP images are decoded in your browser and drawn into an sRGB canvas. The clean copy is encoded as PNG, then optional metadata chunks are removed. Only image structure, pixels, palettes, and transparency remain.
This removes embedded EXIF (including GPS location), XMP, IPTC, color profiles, and embedded C2PA manifests. Displayed orientation and transparency are preserved. Color conversion and bit depth can change, and PNG files can be larger. Animated images are unsupported.
Audio and video: the same encoded tracks
The browser copies supported audio and video tracks into a clean container without re-encoding. Descriptive metadata mappings and chapters are removed. The output container is cleaned again and inspected before a download is offered.
| Format | Removal scope | Preserved |
|---|---|---|
| MP3 | ID3 tags, descriptive tags, cover art | Encoded audio and playback headers |
| M4A / MP4 / MOV | Supported tags, optional metadata and UUID boxes, dates and descriptions | Encoded tracks, timing, orientation, codec configuration |
| WAV | Optional metadata chunks | Format, fact and audio-data chunks |
| FLAC | Comments, artwork, application and other optional metadata blocks | STREAMINFO and encoded audio frames |
| WebM | Tags, attachments, chapters and supported descriptive fields | Encoded tracks and playback structure |
Clean media copies include audio and video tracks. Cover art, chapters, subtitles, and timed-data tracks are omitted. Keep the original if you need those extras. The output checks include track layout, codec, dimensions, audio configuration, orientation, duration, chapters, and descriptive tags.
Audio extraction: a new MP3
The audio extractor is separate from metadata cleaning. It decodes the first supported audio track and re-encodes it as a 192 kbps stereo MP3 at 44.1 kHz. This is a lossy conversion: sound quality, channel layout and file size can change. Video, other audio tracks, subtitles and cover art are not included. The original is not overwritten.
Inputs are supported MP4, MOV, WebM, MP3, M4A, WAV or FLAC files up to 64 MiB and 15 minutes of readable audio. MP3 output is limited to 32 MiB. If a container does not report its selected audio track duration, the container duration is used for the input limit and the output duration is checked separately. Unsupported or encrypted tracks and files without audio are rejected.
Direct HTTPS media-file links work only when the source permits cross-origin browser access and returns a complete file without redirecting or requiring cookies or sign-in. Downloads stop after one minute; extraction stops after two minutes. TikTok, Instagram and YouTube page-link downloading is not connected. Source URLs are never sent into advertising requests.
The extractor checks that the result contains one MP3 audio track, strips supported MP3 tags, and checks output duration and size. Preview and listen to your copy before using it; these checks do not prove that every sample is identical or that all audio content is suitable for reuse.
C2PA and platform AI labels
Use the AI Metadata & C2PA Remover for a focused guide and the same local cleaner.
Supported embedded C2PA manifest containers are removed. This tool does not validate signatures or establish authenticity. External credentials may still be recovered through fingerprints or watermarks, and platforms may label content using other signals. An AI-label bypass is not guaranteed.
See the C2PA Content Credentials specification for the distinction between embedded credentials and recoverable provenance.
References behind the method
The FFmpeg stream-copy documentation explains copying encoded packets without decoding and re-encoding. Quiet Media also applies format-specific container cleanup and output checks; stream copying alone does not remove every kind of metadata.
The C2PA specification describes embedded manifests and approaches for locating associated credentials. Removing an embedded manifest does not remove every possible link to provenance.
How to independently review a clean copy
Inspect the downloaded file with an independent metadata viewer, such as ExifTool, and compare its fields with the original. Check image appearance or media playback as well. The absence of descriptive fields does not prove anonymity or the absence of watermarks, encoded-packet metadata or external records.
Limits and processing time
- Images: up to 20 MiB, no more than 4,096 pixels per side.
- Audio/video: up to 64 MiB input; output must stay within 96 MiB.
- Unsupported or encrypted tracks are rejected. No automatic lossy conversion.
- The preparation display lasts at least three seconds. Larger files and first-time software loading can take longer. Media jobs stop after two minutes.
What remains outside the scope
Required playback data remains. The tool does not scrub metadata inside encoded audio/video packets, change the visible or audible content, remove invisible watermarks, or erase external records.
Method documented September 17, 2026. Browser tests cover supported sample files; device and codec support can vary.