With a Tidal subscription, there was an option to only show server results for Sonic Sage. Without a subscription, that option was not there, and it showed Not Found results like it does now. It seemed backwards since you could not play Tidal tracks without a subscription. Since Tidal is no longer integrated, could the server-only option be added?
The server-only option should be the default (and thus the setting isnât there). You can post Plexamp logs, but it should be searching and trying to match.
When I toggled the server-only option with Tidal, there were no âNot Foundâ tracks. However, the playlist was smaller because they were omitted. This is basically useless now with a three-song playlist.
Log.txt (20.4 KB)
Unless you have those tracks, youâre holding it wrong. The AI doesnât know whatâs in your library, so something like âAcoustic Radiohead tracksâ works well if you have the catalog in your library, but given it doesnât know whatâs in your library, itâs not going to work for those vague descriptions.
I only have those three tracks that it came up with. So, this doesnât work the way I want. Thanks for the explanation, I understand now. The suggested prompts work best if Tidal was still integrated.
Please enable the feature toneed to use Sonic Sage with our personal libraries rather than a music service. The solution is pretty simple: plex creates a csv of all unique âArtist - Albumsâ in your library, gives that to gpt-5-mini as part of the query to create the playlist. Cost would be minimal and GPT-5-mini is smart enough to only use songs from those albums. If you have a 5,000 songs, that is about 417 albums, which turned into a csv is about 4,271 tokens costing only $0.0011 for the input. So if you had 50,000 songs, you are looking at $0.01 for the query.
@elan Yes I just came to write something similar as @timlexi. It is possible to provide the AI with âcontextâ. This context can be any sort of text, doesnât need to be CSV. Maybe itâs possible to provide a text only version of the database to the AI so that it knows which songs it can choose from?
I not often give feedback, but now i think its time for.
I think this Feature (search only local libary) had to be in GPT to activate, but..
better take option âDJsâ for .. (they are not bad so far..)
Filter with reference local? not found â not seen, not counted.. really so strange?ok searches getting to find longer maybe..
so do TIDAL back in Plex !!
wtf - who has done such major sh.. ?
everyone who had it want it again..
plex pass ?- > options gone?
whats next?? no big updates instead same height like Tidal ..
so in time plex minimizes other options again..
thanks Plex - really smart
so, user getting angry.. is normal in this way -
Feedbacks are enough in Forum
Plex is a really good streaming App with Server Musik and Photos. I have the server pass since ⊠long - really good Software at all.
also at Linux and Android/Shield (smb.. is little bit tricky with usb internal..) but ok.
Thanks
yep it isnât complex. And given the user provides their own API keys the cost of tokens is theirâs to wear.
I have done a similar thing in the past but manually (i.e. get text export of everything in my music library) send it to OpenAI API with custom instructions.
Sonic Sage could be an awesome feature for music exploration/discovery in your own collection, but in its current state its not very useful.
TBH, Sonic Sage wasnât that great even when we still had Tidal to fill in the gaps. It would output garbage playlists that didnât match the input prompt at all. I got MUCH better playlists using ChatGPT directly, then importing the playlists into Plex. And this was 2 years ago when AI was still a bit wonky.
Nowadays ChatGPT can output much better playlists, which I love to use for music discovery, but at the moment I have to do that through Apple Music since we donât have Tidal as a backend anymore for tracks that arenât in my Plex library.
I suspect that one could upload a CSV of every artist and album in their Plex library, and have ChatGPT generate a custom âlibrary onlyâ playlist since the tokens and context windows for AI are so much larger now. Worth a try when I have some free timeâŠ
