Wladimir Leite

> @wladimirleite, just when you have time, since you are the original author of this PR, are you OK with last @aberenguel' commits? Sorry about the delay! I checked the...

> ### Shareaza Library1.dat Record > Metadata Field Description > `p2p:sharedFlag` The direct shared flag value from the record. Can be `True`, `False`, or `Unknown`. > `p2p:parentFolderShared` The sharedFlag value...

I'll run a third test processing HDD E01 files, but so far, the proposed change has shown significant improvement. @lfcnassif, do you see any drawbacks?

> Is it possible to get next item randomly in the queue? It's possible, but since it uses a `LinkedList`, removal is O(n). I'll test it, but I suspect processing...

> > Is it possible to get next item randomly in the queue? > > It's possible, but since it uses a `LinkedList`, removal is O(n). I'll test it, but...

Last night, I tested the "queue for high priority / random for the rest" approach. While audio transcription was still slower than during the weekend, it wasn't as bad as...

> Just an observation about this. A key optimization of the transcription service is batch inference, where we process many audios at the same time, speeding up the inference time...

> I am going to repeat the tests this weekend, so we will have a better idea how much each factor is influencing. I repeated the test with case 2...

> Impressive results [@wladimirleite](https://github.com/wladimirleite)! I'm excited about this optimization and waiting for the results with the other UFDRs. Have your test with the E01 evidence finished? Tests with the 3...

> I don't know if this is the same case as your tests, but I noticed that when I processed a ufdr case with many audios and remote transcription, the...