My initial thought was to run the stems through a daw for reducing reverb and turning up a dry version? Haven't tried yet, any other ways to get better results? What would you do. Thanks

 

I'm interested too. Maybe I'd use one of these "frequency/stereo field" isolation plugins to "clean" as much as possible, although I'm not sure if there'd be much improvement. In every stem track before Melodyne, that is. Since sound quality is not important.

I don't know how reverbs affect Melodyne. There's a few "de-reverb" plugins but haven't tried them.

 

well, since you are speaking generically instead of an example; I would use some combination of Accusonnus ERA and Acon Digital plugins on the stem in Logic. You could use RX too, if that is more your style. I'm not the biggest fan of it.

I would probably use also some combination of BlueCat's MB-7 Multiband mixer or Fuse Audio Drums SSX if the stem cleanup would benefit from that.

All of this stuff would be in Logic. So the next step would be to Enable Flex on the track and change to Flex Pitch. Open the Wave in Editor window, and Convert the Flex Pitch Map to Midi. I wouldn't use Melodyne at all for it.

 

important subject. I like to strip some Al Jolson. Time to take the dog out...

 

You are right, in that it shouldn't be important. I wouldn't do any of this first. The only way I would even bother would be if I was getting an unacceptable result. Even "great results" will still only yield so much midi data. It's not going to play back a part perfectly with many included parameters beyond note pitch, note on, types of basic data. It's pretty rudimentary either way, and once you have a decent outline you can easily just draw in the rest.

 

Oh my, I'm still using slowdowner (and it's relatively a new gadget) and ear.
But I admit that tech is so evolved now that it can get 80% or so properly, only, while it's scanning I get already key, bass, chords and melody by ear.
My main problem are the FX soaked details, and here all the above mentioned stuff is really helping.
Cleaning the track from reverbs, choruses, etc. would be immensely relieving for the stem sw.

In anyway, for the most "in mix" arrangement details (and the most difficult thing), scan pan left - right with a steep filter because many voices/instruments are a bit easier to discriminate unless you have mono audio.

 

Right, I hope too.

 

I've tried so many times in each every way mentioned here.

Melodyne , many different deverb algos, stems, it's all good to get you started but if you want a clean transcription you are still going to have to play it at some point. Decoda and slowdowner can help figure out the small details.

If anyone does find a way that they've actually tried that works. I too am curious and willing to give it a go.

 

You can say that again lol. What a pain.
The AI stem splitting is truly one of those "magic today, science tomorrow".