Stem Creation / Exporting Issues in Live 9

[mmr]

Live 9.7.7, Mac OSX 10.13.6, Focusrite 8 Pre X

So I've spent many hours recording and mixing an album and it's time to export the stems.

I've encountered issues with Ableton's export feature before when dithering (Live runs in 32 bit, most deliverable specs require 24 bit) and had gotten into the habit of burning final mixes to a separate audio track in real time. The exported mix using Live's export function just sounded thinner and less full. I did a/b comparisons at the time and after reading many posts online about the same issue, it seemed my ears weren't deceiving me. The magic was simply gone and it sounded way harsh and clinical.

But now I had to create stems. The label required my songs tracks be bounced into 24 bit 48k stems for each instrument. Of course, the same issue occured with the stems too. On checking the stems back the mix wasn't even close to the original. Bouncing them in audio in real time onto a different track as above was not a a viable option as there was silence at the beginning of each audio file and so lining them up manually was far from desirable and was to be time consuming.

After much trial and error, the answer I found was to change the dither option from 'triangular' to 'POW-r-2'. I'm not certain the exported files are exactly the same but they are close.

I am using a lot of external plug-ins (Waves, Slate, iZotope, Soundtools) and I don't know if this has any bearing - but this is what worked for me. I'd love to hear other people's thoughts and input on this.

If I want to burn stems using bounce onto another track in real time (as I was originally doing with my finished mixes) how might I get around the latency issue that happens where it places a small amount of silence at the front of each file?

Hopefully this might also help someone else in a similar position and save them from a few hours of hell.

Lorin

[fishmonkey]

my first thoughts are that meaningful comparisons between very similar bits of audio are very tricky to make, and really require double-blind testing.

if you rule out any mind-bias using double-blind testing, there are a bunch of other things that can inadvertently corrupt your comparison process, including:

- warping
- mismatched loudness levels
- time-varying or randomised instruments or effects that are not 100% reproducible

also, if a null test shows up anything more than very subtle differences between different dithering methods, then it's very likely that there is something else going on apart from the dithering itself...

[mmr]

" - warping"

Do you think it would be better to render all warped audio to new files before exporting? Would this make a difference?

"- time-varying or randomised instruments or effects that are not 100% reproducible"

Would freezing tracks make a difference?

"also, if a null test shows up anything more than very subtle differences between different dithering methods, then it's very likely that there is something else going on apart from the dithering itself..."

I wonder if it's a case of certain third-party plug-ins not playing nicely with Live?

[mmr]

Well I've managed to solve this by bouncing to audio stem tracks within the arrangement.

The export in Live 9 sucks. Even the POW export settings weren't an accurate representation of what was going on in my arrangement. And Ableton support sucks balls too.

[chapelier fou]

I’ve always rendered to 32 bit to avoid dithering.

[mmr]

I was able to avoid the export mode altogether by creating new audio track for each stem then grouping the tracks I needed and sending the group to the new audio channel in record. This solved my issue and the need for using Live’s export function - which didn’t work for me in that no settings gave me a true representation of my mix in Lives arrangement. i wonder why this is?

I’m thinking it’s certain third-party plug-ins aren’t exporting right. Stylus RMX certainly seems to be one of them, the snare hits didn’t sound as full - also guitars - which are using Guitar Rig and Waves plug-ins.

The reason I think this is - once I’d created a new arrangement with the stems and used the export function to export the final mix to check it - it exported just fine… i.e. - when not using any plug-ins.

My question is this.

The label are wanting stems in 24 bit. My audio files are in 24 bit but when using the export function and selecting 24 bit as the option it looks like it is dithering down from 32 bit to 24 bit. If there was no dithering taking place then the dithering options would be blanked out. So Ableton is 32 bit by default? Is this correct? Or am I confused here?