What is nice is the matches for the songs will be what you predict, the matches for the album a bit more broad, however, having things from “Various Artists” now pop up inside of mixes is well worth it imo
You might want to rethink that if you have a lot of “remux” Bluray rips. The non-Pro Shield has trouble with higher bitrates which are usually only occuring on Blurays but not on commercial internet streaming services.
In a mad attempt at cost-cutting they reduced the RAM from 3 GB (Pro) to 2 GB (non-Pro).
Which then made it necessary to use a 32bit Android OS instead of the now usual 64 bit Android to save RAM for the OS…
But even with this measure the usable RAM for apps is more scarce than on the Pro version.
Questions 1 & 2 were answered (thanks Noah0504) but what about #3
I see posts about “personal” stuff and the lack of “libraries” for some platforms…
BUT can I (if not why cant I) take the results to a different system (using the exact same music files).
Sure there may need for a SQL command or so to accommodate different file systems etc.
But once the DB is built it should be “portable” and the results accessible to all systems even if the system playing “this” can’t update it.
I’m currently 8.5% through 3000 albums. Analysis is running on a second server to my server I use for films and TV shows. The music server is an old 3rd Gen Intel i3 3240 with 8Gb RAM and 500Gb out of 4Tb used so far.
I’m contemplating whether or not to leave it running as only I use Plexamp, but if it works on the Roku player as well, I’ll leave it running for what seems like it’s going to take at least a week or more, I may just get another CPU for the board, possibly the i7-3770 which should be a little quicker? is this a rushed out extra for users to test and complain about? Possibly, but it’ll keep getting perfected as time goes on and hopefully quicker?
I do hope this isn’t just a “Gimmick” to get more users onboard like the Live TV side.
As this is machine learning, is there any way it could, at some point, use the HW transcode of users GPU’s to help speed things up? Especially as then the GPU can take the heat from the CPU? or am I going down the wrong hole?
the longer answer is, it may be possible if you cloned your plex data and exact media paths to your other pc, let it process, then cloned it back.
complicating things further is differing windows path vs linux paths (it would be the same issue as moving plex to/from linux/windows).
some one with enough knowledge/experience/patience could maybe pull it off, but it is nothing that is supported and would probably cause more issues than its worth.
alternatively, since plex allows having multiple pms servers on your account, you could create a music library on your PC and use that, removing it from your PI completely.
on plex clients, you would simply pin the windows music server/library and it would show up the same as if the music library was the same server as any other servers content.
it’s definitely no gimmick.
and it works even across shared music libraries (that have of course completed the super sonic analysis).
so, for example, you have your own music library and 2 other shared music libraries, assuming all available music libraries have completed analysis, mixes/radios can use related content from your shared libraries to queue in tracks with your own.
as far as gpu support, I can only imagine it would be as excellent improvement as gpu video transcoding performance became, but only the plex dev’s can determine whether or not it is feasible and worth while to implement.
it could be the type of machine learning required is better suited to cpu, instead of gpu.
I’m admittedly lost with a lot of the technical aspects of this. IT sounds very hardware intensive, If I run this on a qnap TS-453BE with intel celeron J3455 cpu at 1.5GHz and 4GB of memory, is it going to wreck it or something by over loading it?
I currently have 4800 albums in my music library and I’m still adding to it being halfway or so through a music metadata clean up… It’s one thing for it to take a long time, but if I can’t add new music to it until it finishes analyzing and the analyzing is going to take a month or more because my NAS is older, It won’t be worth it. I have my maintenance window set in plex for midnight to 8am. I don’t mind letting it run for a week or so but anything longer will start to be a real pain.
So, it took almost exactly 48 hours. My initial impressions are that the mixes produced are interesting and useful. I would love to see access to them exposed in more than just PlexAmp though…
Anyone has an idea where the results are being stored and how much data this generates? With hughe library I was wonderin if i will run out of space maybe on my partition.
Finished my FLAC library and started to play around with it. There are some interesting results… like Joan Jett matching with Dance Hall Crashers…(Punk/Skacore… ok) while some of my Hawaiian music has no sonic matches.
The mixes are interesting as it has built a mix with some music I’ve haven’t heard in many moons. So, I am gonna give the mixes a spin on a bike ride.
And for those that know and remember… VIVA Radio Free Hawaii!!!
Thanks. Another question. The music analyzer did not restart when the scheduled task window reopened. I know it didn’t complete during the first scheduled tasks window (I have opened the window up to 23 hours temporarily). Should it pick up where it left off automatically and report progress again or do I need to do something to kick start it? The analyzer process is not showing up in my process monitor window nor is it showing up in the plex app progress bar anymore.
Thanks. Manually triggered a music library scan. It ran for about a second and then completed. Looking at the logs now (all ,filter music:). Seeing activity in the logs "music found a match for artist “<>” starting to show up.