Remove Metadata from MKV (Matroska) — Free, No Upload
MKV is the open container of choice for archival, fan-subbed, and multi-track releases. Its metadata is structured as Matroska tags — global Title/Artist/Comment elements plus per-track TagTrackUID + TrackName entries. The cleaner empties all of them and zeroes out track UIDs while preserving the streams bit-for-bit.
Last reviewed 2026-04-28 · Verified against synthetic + real-world MKV (Matroska) samples
Drop your .mkv file
Max file size: ~2 GB (memory permitting)
What leaks in MKV (Matroska) files
- Matroska Title, Artist, Comment, Description tags
- Per-track TrackName and TrackLanguage
- TagTrackUID identifiers (binding tags to specific tracks)
- MuxingApp / WritingApp version strings (Lavf, MKVToolNix, etc.)
- Encoder hint strings inside the Tracks element
Confirming it worked
| Check track names, not just the header | MKV stores a title per track as well as globally. A file that looks clean at the top can still name the ripping software on an audio track. |
|---|---|
| Chapters and tags | MKV carries chapter names and arbitrary tag elements that other containers do not. Confirm both are gone in the viewer. |
| Attachments | MKV can embed fonts and cover images as attachments. Those are separate from metadata; check whether the cleaned file still carries any. |
| What it does not remove | Subtitle tracks, which are content rather than metadata, and which frequently contain a release group name in the first line. |
Frequently asked questions
How do I remove tags from an MKV file?
Drop the MKV in. The cleaner writes out a new Matroska container with all Tag elements emptied and track UIDs zeroed. No re-encoding.
Does it preserve subtitle and audio tracks?
Yes. `-map 0 -c copy` keeps every track (video, audio, subtitles, attachments) bit-identical. Only the metadata wrappers around them are emptied.
What about MKV chapters?
Chapter lists are removed (`-map_chapters -1`). If you need to preserve chapters, this tool isn't the right fit — chapter timestamps + names are themselves identifying metadata.