People misidentified & NFO is correct, yet Emby's metadata is uncorrectable.

[rexerm 36]

I see this happen constantly with same-named people and it is driving me nuts. Because Emby is so unreliable for identifying people correctly, I now manually add precise IMDB and TMDB IDs to my NFO files. This still does not help in all cases, even with zero Metadata Downloaders active for these libraries (just NFO-reading turned on). For example, Eric Johnson has metadata in Emby as in metadata1.jpg below. I then scan the series Smallville, which contains this in its NFOs:

<actor>
  <name>Eric Johnson</name>
  <role>Whitney Fordman</role>
  <type>Actor</type>
  <tmdbid>33337</tmdbid>
  <imdbid>nm0425053</imdbid>
</actor>

This results in metadata2.jpg, in which Emby has -- for no reason -- changed the IMDB ID to nm0425054 (the wrong Eric Johnson), but left the TMDB alone. I then scan the movie Drive, He Said, which contains this in its NFO:

<actor>
  <name>Eric Johnson</name>
  <role>PFC. Johnson</role>
  <type>Actor</type>
  <tmdbid>4884201</tmdbid>
  <imdbid>nm0425054</imdbid>
</actor>

This results in metadata3.jpg. Now this older Eric Johnson's TMDB and IMDB IDs have completely taken over for the the younger actor from Smallville. I can remove both from Emby, scan, add them both back, scan, and get these same results. I can manually edit the IDs in Emby for this person, but if I rescan Drive He, Said, it introduces the errors again.

There seems to be no way to get Emby to treat these as two people. Why is the NFO data not respected for people?

[Luke 42077]

Quote

This results in metadata2.jpg, in which Emby has -- for no reason -- changed the IMDB ID to nm0425054 (the wrong Eric Johnson), but left the TMDB alone. I then scan the movie Drive,

It could have come from internet metadata, which may have still been on even though you thought it was off. The server log from when the series was first added would help tell us that.

If that doesn't lead to anything, then we'd have to know what entries for Eric Johnson were in your server database prior to adding the series because the server will try to find the existing person in your database before adding a new one.

[rexerm 36]

Looking in metadata\people folder, there are these three Eric Johnsons, with the IDs in the person.nfos as follows:

Eric Johnson-imdb-nm0425053
  <imdbid>nm0425053</imdbid>
  <uniqueid type="imdb">nm0425053</uniqueid>

Eric Johnson-tmdb-33337
  <uniqueid type="tmdb">33337</uniqueid>
  <uniqueid type="imdb">nm0425053</uniqueid>

Eric Johnson-tmdb-4884201
  <uniqueid type="tmdb">4884201</uniqueid>
  <uniqueid type="imdb">nm0425053</uniqueid>

Which... none of these match what the current Eric Johnson profile has in it (4884201 and nm0425054), but the top two both match the Smallville actor and the date in my NFO. Presumably, nm0425054 is being pulled from emby's Db?

Attached is the server log. I did the following:
11:58pm - Removed Smallville and Drive, He Said from their libraries and rescanned both. Corrected Eric Johnson with IDs nm0425053 and 33337.
12:00am - Added back Smallville, rescanned "Episodes" library. Eric Johnson has now changed to nm0425053 and 4884201.
12:02am - Added back Drive, He Said, rescanned "To Watch Movies" library. Eric Johnson is now fully converted to nm0425054 and 4884201.

Note: sometimes I observe Emby editing all NFOs when data for a person is changed in the metadata editor. In this case, it is not doing that and both NFOs remain with their respective (different) Eric Johnson data.

embyserver (1).txt

[Luke 42077]

What's more important is what was in your database prior to adding the series.

The good news is that I don't see any moviedb or tvdb requests in this log, although I also don't see any references to Smallville either, so I can't be sure that this is the log from when the series was added.

[rexerm 36]

Is there a way to "clear out the cobwebs"? That is, can I erase all people data, then rescan my entire set of libraries again to rebuilt them from my NFO data? I know I could delete the folder in metadata easily enough, but that's, if I remember correctly, only data for people created or altered since that feature became active.

[Luke 42077]

Let's assume that what you said about having had internet metadata already disabled is true. So we can rule that out. Then:

Quote

This results in metadata2.jpg, in which Emby has -- for no reason -- changed the IMDB ID to nm0425054 (the wrong Eric Johnson), but left the TMDB alone. I then scan the movie Drive, He Said, which contains this in its NFO:

Most likely you already had Eric Johnson in your server database containing the same tmdb id but the different imdb id, and the server linked to that one rather than creating another one.

So what I would do is focus less on the person object and focus more on all of the content that the Eric Johnson's are linked to, such as the movies, series, episodes, trailers, etc. I would go through the metadata editor for all of those and make sure they look correct in the cast list.

[Luke 42077]

Ah but actually I forgot. We have not yet added external id's to that screen. So you might have to do this by looking at your nfo files and then running a normal library scan following any changes.

[rexerm 36]

An "Eric Johnson" is present in two other shows (currently linked to this same profile). However, those two shows are ones that I have not gotten to yet in my "update everything with IDs" project. So, those NFOs contain just:
  <actor>
    <name>Eric Johnson</name>
    <role>Guest Star</role>
    <type>Actor</type>
  </actor>

So, yes, likely there's all sorts of old people data cached from years of running this server. My main concern is: I've found this one particular person's error... likely there are many more. Not only does there not seem to be a way to fix this one, I don't see a reliable way to identify any others. Currently, I am able to scan all NFOs for changes as compared to my own personal database of metadata -- a source of truth I use to construct the NFOs in the first place. However, the NFOs are not being altered in this case... only emby's internal metadata is changed.

Going back a bit, I have these two entries in the metadata\people folder:

Eric Johnson-tmdb-33337
  <uniqueid type="tmdb">33337</uniqueid>
  <uniqueid type="imdb">nm0425053</uniqueid>

Eric Johnson-tmdb-4884201
  <uniqueid type="tmdb">4884201</uniqueid>
  <uniqueid type="imdb">nm0425053</uniqueid>

When Smallville is scanned, it seems to be identifying that actor with the Eric Johnson-tmdb-4884201 entry, despite my NFO data exactly matching the  Eric Johnson-tmdb-33337 entry. Why would that happen? (also, the log did contain Smallville entries, starting at 2025-12-01 00:01:09.041)

[Luke 42077]

Quote

So, those NFOs contain just:
  <actor>
    <name>Eric Johnson</name>
    <role>Guest Star</role>
    <type>Actor</type>
  </actor>

Possible culprit here. When the server looks for eric johnson in your database, if it can't find any by id it will look by name.

[Luke 42077]

Quote

Going back a bit, I have these two entries in the metadata\people folder:

Again, not as important as the data that is attached to the media directly. I would clean that up first.

I didn't want to say this earlier but now that there's evidence of it, I will just let it out. If you let Emby Server manage your metadata without pre-existing nfo created by other software, you will not run into this kind of thing.  

It is only when you bring messy data into the mix that this happens.

[rexerm 36]

No, it was not messy data. Removing the two shows mentioned above (that lacked any IDs in the NFOs) didn't change anything upon re-testing. However, I just moved the Drive, He Said movie to a different library and scanned it there. This time, it left Smallville Eric's profile alone and created a new Eric with its own IDs, as it should.

My guess as to the difference: "To Watch Movies," in the past, used to have 1) no NFO usage and 2) would scrape TMDB for data. Those were both changed, but perhaps there were some remnants of this scraped data persisting somewhere? I think i will re-create this library with the newer settings from the start.

I will say: NFO support is one of the strong points for Emby for me vs the competition. I love being able to precisely define the data, rather than relying on scraping the sometimes incomplete or inaccurate data TMDB has.

[rexerm 36]

Related: there is no easy fix for this situation, since there is no merge ability for people profiles. This man has the exact same IMDB and TMDB IDs in both profiles, but his name happened to be spelled differently once for some movie I imported long ago. So, Emby dutifully create an entirely separate profile for him, rather than merging it with the existing profile with the identical IDs.

I would like to see Emby prioritize ID matching, rather than name matching. Name matching is clearly prone to error, particularly with names like those from Asia which can have a lot of variety in spelling (or even last name / first name order) when translated to English.

[Luke 42077]

Hi, id matching is the priority, as you’ve already learned in the previous examples.

[rexerm 36]

I have not learned that. If ID matching is prioritized over name matching, why did Emby create a duplicate person entry that had identical IDs and mismatched names? In the previous example, Emby mixed up two same-named actors, going so far as to alter IDs to match one to the other. And, in the other thread I made for a similar issue, Emby is altering IDs for same-named actors directly in the NFO when unrelated metadata is changed.

Additionally, the fix for when these errors crop up is arduous: removing all related media from all libraries, scanning, adding the media back, rescanning. Sometimes, as above, it requires re-creating a library entirely. Why can't a simple scan detect and repair these issues for actors/directors/writers?