Hi, thanks for doing this again.
The quality of the metadata in Plex’s database (and thus the database used by the new scraper) is atrocious. I made a few dozen threads in the metadata forum. There are weird nonsense AI descriptions. Default poster suggestions are across the board bad. Several movies seem to be subject to copyright claims and have no default posters. Release dates are weird all over the place – I reported several dozen that were off by more than 5 years but I have literally hundreds that are off by a year or two.
Some of these issues have been fixed silently in the intervening period. I got a reply from moderators once or twice. Lots have been completely ignored. Just talking about box art, here are five unresolved issues (I have another 5 or 6 that are also unresolved for box art, and then the other sets of issues above):
Since I didn’t get replies on any of these, I have generally stopped reporting the issues. I went ahead and fixed literally thousands of posters in my own library as a result of the switchover to the new library agent.
I think the thing that’s frustrating here is:
- All of this stuff was fine with the old agents. I’m aware of the fact that a combination of the streaming services’ needs, copyright risk, changing data license models, etc. led you guys to make your own database and use Gracenote(?) for covers, but the result is really significantly worse.
- In addition to the tons of actual weird junk posters and crappy, poorly composited posters, even “good” posters are bad: almost all of them feature huge closeups of an actor and many of them have actor names and other text on the posters. I get that this is Gracenote’s decision, and I am sure an algorithm was designed to maximize engagement or maximize some other KPI but it’s really at the expense of poster quality. Most people prefer relatively clean posters, and prefer the use of director/studio-approved key art rather than just random shots of actors.
- Despite things getting worse on the metadata side, I’m happy to fix issues myself. But unlike the old agents, I can’t go into the source data and fix the issues because you don’t have any public or API backend for your data that’s user-editable. So I can’t fix anything myself.
- That’s fine, I’m willing to report issues to you so your staff can fix the issues, but it seems like you don’t have the staffing bandwidth to do so or don’t agree these are issues.
- Frustratingly, there appear to be several users here in my boat who volunteer their time writing scripts to automatically surface and report metadata issues – anything we can do for free, Plex is better situated to assign paid staff to. This is a perfect job for something like a GSOC project; developing heuristics and rules to automatically flag low quality metadata.
- No other product that uses movie or TV metadata suffers from these issues right now, just Plex.
It feels a little bit like metadata is something you’re reluctantly doing because it’s necessary, not something anyone on the team has any real passion about. It seems underdocumented and understaffed. That’s a real bummer. The scrapers and the library metadata are, after video playback, the single most core functionality of the software.
All of this combines to make me have very little faith in Plex’s efforts to dive DEEPER into managing metadata, for example with the currently-in-preview user review function. If you can’t handle movie level metadata, moderating user reviews is going to be an absolute nightmare. I have experience moderating multi-hundred thousand user websites and my day job is text-as-data for user generated content online and you guys are not set up for success if you choose to take this on.
Please work on improving metadata quality in an actionable way (not just “it’s a priority for us”).
