Idea: Audio Fingerprinting Plugin for Media Identification (Shazam for Jellyfin?)

[hrastovc]

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:

1. 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).
   
2. Crowdsourced Database: Since no massive global database for movie/TV audio fingerprints exists, it would rely on a crowdsourced model.
   
3. 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!

[TheDreadPirate]

While I am not against this, in principal, I feel like the MUCH easier task would be to name files properly.  There are tools to automate renaming of files, including to make them conform to Jellyfin's documented expectations.

Regarding feasibility, consider Musicbrainz.  It is run by the Metabrainz Foundation and has several PAID employees, plus all the volunteers that contribute to their code base AND the thousands and thousands of volunteers and users that contribute the fingerprints and metadata that power Musicbrainz.  Metabrainz spends hundreds of thousands of dollars a year for employee salaries and for hosting costs.  Any kind of movie fingerprinting DB would also incur significant hosting costs.

I can't see how this would be feasible without significant fundraising and an organization, similar to Metabrainz, being stood up to drive this forward and keep things on track.

I'm not saying it can't or shouldn't be done.  But this is not something that can feasibly be done in your free time since the time commitment to get the database to a size where it is useful is not insignificant.