2026-06-13, 04:33 PM
Hi everyone,
I've been thinking about a persistent pain point in media self-hosting: media identification for poorly named or obscure files. Right now, we rely heavily on strict file naming conventions for TMDB/TVDB scrapers, or file hashing for things like subtitles. But if a filename is completely butchered, or if a video file has been remuxed or altered slightly, traditional matching often breaks down.
I’m wondering if anyone has ever explored, or if there would be interest in, a Jellyfin plugin that identifies video files using audio fingerprinting, similar to how Shazam works for music or how AcoustID operates in the MusicBrainz ecosystem.
How it could work:
Why do this?
The Challenges:
I know there are massive hurdles here: dealing with different audio dubs/languages, different audio channels (5.1 vs stereo), the processing overhead of scanning full-length movies, and the "cold start" problem of building a database from scratch.
Before I dive too deep into the technical feasibility or look into writing code, I wanted to throw this out to the community:
Would love to hear your thoughts, critiques, or if anyone would actually find this useful!
I've been thinking about a persistent pain point in media self-hosting: media identification for poorly named or obscure files. Right now, we rely heavily on strict file naming conventions for TMDB/TVDB scrapers, or file hashing for things like subtitles. But if a filename is completely butchered, or if a video file has been remuxed or altered slightly, traditional matching often breaks down.
I’m wondering if anyone has ever explored, or if there would be interest in, a Jellyfin plugin that identifies video files using audio fingerprinting, similar to how Shazam works for music or how AcoustID operates in the MusicBrainz ecosystem.
How it could work:
- Local Scanning: The plugin runs a background task to generate an acoustic fingerprint of a video's audio track (perhaps using an algorithm like Chromaprint, similar to how the Intro Skipper plugin works locally).
- Crowdsourced Database: Since no massive global database for movie/TV audio fingerprints exists, it would rely on a crowdsourced model.
- Confidence Matching: When multiple users with properly matched libraries scan the same file and upload the identical audio fingerprint, the system links that fingerprint to a specific TMDB/IMDb ID with high confidence. Over time, a global lookup database is built.
Why do this?
- It bypasses bad file names completely.
- It’s more resilient than standard file hashing (which breaks if you strip an audio track, change a container, or edit metadata).
- It could help identify rare, obscure, or home-archived media that traditional scrapers miss.
The Challenges:
I know there are massive hurdles here: dealing with different audio dubs/languages, different audio channels (5.1 vs stereo), the processing overhead of scanning full-length movies, and the "cold start" problem of building a database from scratch.
Before I dive too deep into the technical feasibility or look into writing code, I wanted to throw this out to the community:
- Has this been tried or discussed before for video media?
- Do you think the community would have an interest in contributing their library signatures to a shared database like this?
- Are there glaring technical roadblocks I might be overlooking?
Would love to hear your thoughts, critiques, or if anyone would actually find this useful!


