Offline Bounce vs Realtime Bounce vs Recorded-in-Session - BIG differences in PT2021

[andresix]

Hi everybody !

Just realised the other day while comparing my exported mix to what I had in the session that they didn't sound exactly the same.
At first I thought it must be me, but after a couple tests, it became obvious that the OFFLINE bounce just wasn't the same as what I was hearing during mixing or what came out of a REALTIME bounce.

I'm on PT 2021.3.1, not HDX, 2018 iMac on macOS Mojave (10.14.6), running only native stuff.

When doing a null tests, there's a lot of Side stuff missing from the offline bounce, a bunch of high end sizzle and things like that.
All the tracks end up in the mixbus and my session is structured with a few bus cascades but nothing super complex.

I did a bunch of tests :
- offline bounce (using physical output as source)
- offline bounce (using a bus as source)
- realtime bounce (physical output)
- recorded in session (bus to audio track)
- commit of my mixbus (offline)

None of those pass the null test with none of the other ones... I imagine some of it is just randomisation of my effects or something like that but I've done this test in a couple of sessions and the offline bounce is always a bit darker, with a smeared stereo image and lacking transient precision.

Here are some of those tests to listen to if you're interested.

Anybody else experienced this ? I have been having this impression of lack of detail and precision for a while now, with all of my projects but I've always thought that it's just me, overthinking it.

It would mean that I've been sending lesser versions of my mixes to mastering for months/years. By the time I bounce the final version I don't really think anymore to be that critical, I kinda assumed Pro Tools won't screw up my work. How naive .


I would love to find out that I'm making some mistake, so challenge me !

[GregX]

I can hear the difference in your files. I started getting problems with offline bounce lately, where the bounce was a minute longer than the actual song, filled with blank space I hadn't selected. Also, I was watching a Groove3 course today, Vocal Production Masterclass, and the instructor, Marcus Huyskens, says that he doesn't use offline bounce because he likes to take the opportunity to hear the song one last time to check for any problems. He was not even using Pro Tools.

So, based upon your experience and excellent analysis, my current song 'length' experience, and the Groove3 class lessons, I will plan to bounce in realtime from now on. By the way, your song was great, I was looking forward to hearing more of it when the selection stopped.

[Darryl Ramm]

Thanks for providing the null tests, interesting.

What plugins are you using on this session?

Can you share the session?

[JCBigler]

Originally Posted by GregX

and the instructor, Marcus Huyskens, says that he doesn't use offline bounce because he likes to take the opportunity to hear the song one last time to check for any problems.

That's great when you only do 3 1/2 minute long pop radio mixes.

When you're working on long form stuff like full symphonies and operas, it's a royal PITA to wait for a full bounce to render and then have to go back and re do them after a 2 or 3 hours long recording. Eats up a tone of time and leaves your computer tied up for a half a day, or day at a time. That blows.

[Dizzi45Z]

I know specifically one example that consistently does not match between Faster than Real Time Bounces and Real Time Bounces:

Pro Tools Inaccuracy in FTRT Bouncing vs Real Time Bouncing

This example shows that if you use a mono/stereo plugin on an Aux Input and make it inactive and then use stereo plugins afterwards, it will not accurately bounce the audio that goes through it.

This has been a bug for a very long time and almost caused a huge issue for me on an important album. I was swapping out the reverbs on Aux tracks consistently. Luckily, I noticed the difference before the album was finalized.

However, I must say that besides this one issue, I render faster than real time all the time and I haven't noticed any other weird FTRT bounce issues besides this one.

[Frank Kruse]

I just had a case where I could not get the rendered version of elastic audio sound the same as the real time EA. The ONLY way I was able to print what I was actually hearing was real-time re-recording to another track.

Offline bounce was completely different in both sound quality and sync, real time bounce again different. The only way to preserve the EA work was to real-time re-record into another track within the session.

Also tried switching real time Rhythmic to rendered processing, resulting in different sound. Also X-Form extremely different in a bad way.

All I wanted was to print what I was hearing from the real time "Rhythmic" mode.

Also: Pulling the EA audio clips to a non-EA track to force the render sounded completely different than the realtime track in both sync and sound.

So EA could also be a reason....

But in general I only use offline bounce for temp guides, never for actual masters as I have to listen through them anyway for QC so the time saving aspect doesn't apply plus you can update with destructive punch which is MUCH faster than another offline bounce of the entire thing.

[XJENSEN]

Yes, I can confirm the problems with mono/stereo plugins. Offline bounce cannot be trusted, sorry.

[underground123]

I have often had issues with "offline" bounce ever since it was introduced in Pro Tools. I made the decision to only do real time bounces or record to a new audio track through a buss.

I know that doesn't help, especially with long program material but unfortunately I just don't have confidence in the offline bounce system used in Pro Tools.

Side note: Track commit also does NOT null depending on the plugin. Reverb and other randomized plugins excluded from the null test.

Just sayin'

[smurfyou]

Originally Posted by Frank Kruse

I just had a case where I could not get the rendered version of elastic audio sound the same as the real time EA. The ONLY way I was able to print what I was actually hearing was real-time re-recording to another track.

Offline bounce was completely different in both sound quality and sync, real time bounce again different. The only way to preserve the EA work was to real-time re-record into another track within the session.

Also tried switching real time Rhythmic to rendered processing, resulting in different sound. Also X-Form extremely different in a bad way.

All I wanted was to print what I was hearing from the real time "Rhythmic" mode.

Also: Pulling the EA audio clips to a non-EA track to force the render sounded completely different than the realtime track in both sync and sound.

So EA could also be a reason....

But in general I only use offline bounce for temp guides, never for actual masters as I have to listen through them anyway for QC so the time saving aspect doesn't apply plus you can update with destructive punch which is MUCH faster than another offline bounce of the entire thing.

I really dislike Elastic Audio for many reasons, this is one of them. Was your source a WAV file or MXF? I've noticed issues randomly when it's MXF. Doing a consolidate to force a new WAV file clears it up. Then it sounds as expected, real-time or rendered.

[Dizzi45Z]

Originally Posted by andresix

Hi everybody !

Just realised the other day while comparing my exported mix to what I had in the session that they didn't sound exactly the same.
At first I thought it must be me, but after a couple tests, it became obvious that the OFFLINE bounce just wasn't the same as what I was hearing during mixing or what came out of a REALTIME bounce.

I'm on PT 2021.3.1, not HDX, 2018 iMac on macOS Mojave (10.14.6), running only native stuff.

When doing a null tests, there's a lot of Side stuff missing from the offline bounce, a bunch of high end sizzle and things like that.
All the tracks end up in the mixbus and my session is structured with a few bus cascades but nothing super complex.

I did a bunch of tests :
- offline bounce (using physical output as source)
- offline bounce (using a bus as source)
- realtime bounce (physical output)
- recorded in session (bus to audio track)
- commit of my mixbus (offline)

None of those pass the null test with none of the other ones... I imagine some of it is just randomisation of my effects or something like that but I've done this test in a couple of sessions and the offline bounce is always a bit darker, with a smeared stereo image and lacking transient precision.

Here are some of those tests to listen to if you're interested.

Anybody else experienced this ? I have been having this impression of lack of detail and precision for a while now, with all of my projects but I've always thought that it's just me, overthinking it.

It would mean that I've been sending lesser versions of my mixes to mastering for months/years. By the time I bounce the final version I don't really think anymore to be that critical, I kinda assumed Pro Tools won't screw up my work. How naive .


I would love to find out that I'm making some mistake, so challenge me !

By the way, I still think the issue is to do with EFX Bussing with mono/stereo plug-ins (see my post above). I downloaded your Sophie - Text Export - No FX - Offline Main Out and Sophie - Text Export - No FX Realtime Main Out and when I flip the phase on one, they null completely.