I confirm as a plugin developer that we utilize random noise, tolerance modeling (allow margin of error) etc... so yes, no two bounces will sound 100% the same.
Thank you Ziad, and thank you for your email - we’ll be in touch in the next couple of days!
@@PresentDayProduction Cool! Does that mean that the differences we hear in the null tests are due to those randomized and modeling elements?
@@PresentDayProduction Thanks so much! I hadn't had a chance to view your entire clip (will complete soon) so forgive me if it's addressed there.
It’s in there :p but now you’ll feel twice as pro because you already know what’s coming 😎
Thanks for this detailed analysis! I can add two things to this (I am a software engineer working on Logic Pro in now my 30th year): (a) our plugins do have randomness, we just take great care that the randomness is _not_ screwing up reproducibility. For exactly the bounce reason: you want the identical bounce every single time! It also has advantages for software testing, e.g. when we were working on the Apple Silicon version of Logic Pro last year. We obviously wanted that version to generate identical bounces on the same input - but with a completely new CPU and new tools, that is not necessarily guaranteed. Thanks to a/b testing we were sure that they are identical (for all our test cases). The realtime bounce has once specific purpose, which today is mostly an outdated reason: if you have external MIDI hardware, which is part of your mix and you want to record that audio output into your mix, the bounce _has_ to be realtime, because the hardware simply can't be driven faster than realtime. If you are only using software-only plugins, there is NO reason to not always do an offline bounce - the math is not different from the CPU calculating at 44.1kHz or as fast as possible, if the result is written into an audio file anyway.
Thank you for your comment Markus! We’ll include that in our next ‘news’ episode, that’s interesting to know. Thank you for posting!
How do you guys on the Logic team make the randomness reproducible? My guess is to use the same seed? Is that correct?
When you mentioned “outdated” in reference to using offline bouncing for MIDI in logic pro did you mean that it now uses a similar operation to “Freeze tracks” that they use in digital performer? (I’m a old DP user re-familiarizing myself from years off.)
As someone who recently finished an album of offline bouncing, what a relief haha
I am so impressed with how you went about this. Every other channel I’ve seen tackle this has performed a single null test and decided “real-time bounces are better.” You went deeper, consulted an authority, and then did another null test to possibly disprove your conclusion. This evinces a better understanding of the scientific method than most folks seem to have. Bravo!
Now when my urologist tells me my mix sucks, I can just tell him it’s Apple’s fault cuz my Logic compressors aren’t random. 😬
I haven't even watched the video yet, but your comment has left me stunned.
Ha! My urologist happens to be an amateur musician who also uses Logic. Because I have a tendency to pass kidney stones a lot more often than the average, he sees me every quarter and we always talk about the environment or new features/tricks we’ve learned.
@@russellszabadosaka5-pindin849 My eyes are starting to lock open, reading this thread. Oh, it's 9am. Time for bed.
@@russellszabadosaka5-pindin849 Ha! That’s awesome. My urologist saved my left kidney so I’m no longer contemplating suicide as an escape from pain.
Hallelujah! I can continue to save time by not having to listen to my mediocre mix again... 🙉
Years ago, when I was studying audio engineering, our teacher said that we should always bounce in real-time. I argued that it wouldn't matter, but he replied that he performed THE TESTS HIMSELF. I tried to say that if he used any sort of analog modeling plugins, then of course it wouldn't pass the null test, as these are made to be random, but he wouldn't listen and basically shushed me, lol. Mind you, he was (maybe still is) one of the major certified Cubase trainers here in Russia.
Thank you. Common sense would tell you the computer will calculate the algorithms the same way every time. This was an exceptionally succinct proof of your premise. I have long bounced offline as a way to mitigate an overloaded processor near the end of a big production, the idea being that offline the computer has time to perform every calculation perfectly with full resources. Your demonstration has confirmed my own conclusions quite nicely here. 😊
you guys are awesome. lol the UAD comment at the end gave me a good laugh.
Thank you for the great videos! Keep them coming!
Insanely well thought out test, not only using logic as a DAW, but as a basic modus operandum! You guys rock!!!
This is something that I'd been thinking about recently in Cubase so it's pretty fortuitous that this popped up in my suggestions. Great video. Subbed.
Thank you for digging into these fascinating details!
Thanks, very informative. Good to check which plugin's bring some randomness to game!
The presentation on this channel always reminds me of those late-80s/early-90s science lesson videos 😂
That first switching example though, there is absolutely a difference. Closed my eyes and could hear the
schanges. It's subtle; the real-time is slightly 'girthier'. Not really sure how to explain it. It's certainly slightly
wider, but it's not just a width thing.
You can set some plugins to perform differently like having higher over sampling when bouncing. Offline can give your computer time to process what it can’t in real time.
Very good point, in this case the offline render would actually be slower, but in theory should sound better because there actually is a changing factor.
FANTASTIC! Thank you guys!
You guys are great! You did every test I wanted you to. I was so ready to comment with “but you should have done this..” but by the end I was completely satisfied.
Great to hear, thank you for your comment! Always nice to see a comment saying we covered everything required!
Great site which dismantles with down to earth rigorous precision many of the audio myths that are disseminated by those on the internet who really shouldn't be allowed within a million miles of a web cam and a microphone. Keep up the good work fellas.
Really interesting! Thanks for creating a video about this , with comparing results.
Love you guys! The relaxed vibe is a nice extra to watching these interesting comparison.
Debate settled! I have always argued that the "differences" have to do with processing. Now I know I was right. Thank you.
Another great video gentlemen. Thanks for the info
Just spent 10 minutes watching this video and that actually saved me 20 minutes of real time bounce. So it was definitely worth it. Thanks!
I absolutely loved this video. Thanks so much! Thoughtful, clear, thorough, and definitive. Just shared to my facebook :)
exactly what i thought. thanks for this, i always enjoy your topics.
Such a useful video, thank you!
Great video, thanks!
Really great video!
just finished our album and bounced offline, and panicked a bit watching this.. phew!
First time on this channel. For the first half of the test I felt like shouting at the screen "You need a control where you do two bounces of each and null test them!" until you did exactly that. I noticed a while ago that I could do 2 bounces of the same mix and sometimes get a difference of up to 0.5dB peak level but couldn't hear much audible difference so did a null test and put it down to algorithmic reverbs and analogue modelling but wasn't sure. Thanks for putting my mind at ease. For me the big advantage of offline bouncing is the ability to export a mix with a cpu load that wouldn't play back, that way before a final bounce you can use maximum oversampling on any plugs that use it. It can start to take longer than a real time bounce but I have noticed a subtle difference in clarity when A/B testing those bounces. Subscribed.
Thx for clearing this up🔥🤘
Hey!
very interesting topic you bring up here! This was a topic back in time that was unclear when Protools refused to process its bounces while Logic did so.
I want to clarify that there have been differences in large sessions with extremely advanced editing on edit when I drove a Power mac G4 loaded with the highest frame memory that went at that time with a Digi-001 sound card and extra converter of Alesis AI3. I had a major recording and mix for an upcoming Reggae band and on 2 different occasions there were unwanted changes to their bounce when we processed their bounce and were to be sent to the mastering studio. An occasion when one of the songs got less deley on the verse song. The singer contacted us and pointed out the change from the previous mix. The customer was right, changes to automation had been made as we had bounced "processed" this song at this very moment!
Another occasion was a ballad, the singer had full control of her wishes and sound, a change in the processed bounce had taken place when a "mute" on the dubbing of the voice had been deactivated and a digital snap of a fade had been dropped and a digital snap came with and process bounce!
After these experiences back in 2002, I always ran a real-time bounce to have full control of what happened from the computer process. Previously, we ran a full analog studio with 24 channel tapes and an old MCI 800 table. My conclusion in my post is the following!
If you do NOT have a stabilized current in your studio, an older or smaller daror with weak processing power, there may be changes in the processed bounce! This from my own experience back in time! Love your videos and your topics you cover! wonderful to follow you, you deserve incredibly more viewers! Happy Easter from Carl-Magnus Consitt!
Will save me some time! I was talking to the screen asking, what about if you tested two of the same mixdown type, 2 minutes later you did it. Awesome job.
Great video thanks
Another source of difference might come from oversampling. Some (smart) developers turn on oversampling when rendering offline, and that's usually for the better. Sometimes that might be undesired though, in particular on compressor and distortion plugins because the final result might sound drastically different (for crazy distortion settings, sometimes the aliasing you get from running it without oversampling might get you the tone you were after). Awesome video as usual guys
You guys just blew my mind a bit. Instant sub. I feel sick 😂
Always wondered this! Thankyou
Very interesting 👍 thanks for the video
Thanks for this interesting topic!
Thanks for this!
Brilliant. Brilliant. Well Researched and Superbly demonstrated. Good Job.
Very interesting, thank-you!
I smiled at the notion of waiting whilst a valve emulation plug-in warms up 😂 I bet Wytse would call that Snake Oil!
Ha! That was fun! Thanks guys
Well made and informative
Just wanted to say thanks for the content folks, y'all have helped me out a bit .
This was driving me crazy and you just answered it. Thank you! wasn't aware randomisation really is a thing.
Good to know... thanks for the video.
VERY interesting. Thank you 🎧🎧🎧
Very interesting, thank you
Compliments for the very clear video!
Very impressive information
Well done ! In the first few minutes I thought "yeah, well but did you certify it with two different renders of the same origin?" - and you just did that. Insightful on how much variance is introduced with hardware modeling. Should try this out with my own plugins...
Excellent. Thx
I love the intro to this one SO MUCH.
And also, KILLER video in general! I hadn’t even considered this before and I always just assumed they were the same. Obviously they are, but hadn’t considered at all how the plugins are designed.
Thank you! It will be available on all major streaming services, cassette, vinyl and wax cylinder very soon. I’m also available for session work, weddings and bah-mitzvahs should you be interested 🤣
Just love these dudes !!
Lovely vocals, Mark!!! ❤️
Thanks, Javier! I did warn you earlier.... 🤣🤣 BVs needed on your track???
Super interesting! Learn something new every day. :)
Very good!
Interesting. Cool video!
Great content on this video and great channel! Thank you! I stumbled across this video because I was looking for bug reports on differences between what you hear inside the project and the "printed" file, because I had this really big issue with one particular project recently. It seems like one or more plugins don't work ONLY when bouncing, but there's no message, warning, cpu crash... Have you ever been through anything similar?
First time viewer here. Good stuff guys. I like the way you approach things.
The intro is hilarious - one of your best videos yet. What about uad plugins in reaper? 😂
I love your channel 🥰 Change name to Audio Mythbusters 😄
That Intro will DETUNE an AUTOTUNE
Years ago (though I don't know when it changed) offline bouncing would screw up anything with sidechain compression. It wouldn't follow it properly and the timing would be off on the 'pumping'. So I kept waiting for you guys to talk about that, and you never did. So I went right to my computer, put in a few loops, added sidechain from a kick and did my bounces. Then brought them back into Logic, reversed phase, and viola, silence! I'm a believer now! So for final mixes, I still like to listen as it's going down so I'll go realtime. (old analog habits) But for rough mixes now, OFFLINE! Great demo, and thanks, you've saved me time, 123!!! 👍
love the intro about the scientists lol
What I've heard on occasion and done the test several times for clients who have never had the conversation settled, I've noticed that the stereo field slightly (like as in microns of difference) collapses on offline bounces. I've done this on several DAWs over 20+years, I remember when Cubase and Digital Performer started "offline" bounces I would sometimes on very rare occasions could tell something would happen either defined as M/S or Stereo. I literally used to chalk this up as the "random algorithmic" characteristics of plug-ins. So I guess in conclusion that its so small of a difference its negligible. That being said, my preferred way of exporting tracks is printing out through my Lunch Box 500 which contains a pair of TubeTech EM-1A and SSL Bus Compressor. No matter where I go I always get "That Sound".
I really like that song. It has an early 70's feel!
I have been bouncing in real-time for a while now. Mostly because in Cubase, certain VST instruments would not always bounce properly when done offline. It hasn't been an issue with real-time bounces, so I just stick with that for reliability.
Virtual instruments on their own are not accurate. I always bounce them into audio before mixing.
Great video, gents!
But what about UAD plugins in reaper?
@@PresentDayProduction Even funnier because for the most part I just use Logic stock plugins. I just couldn’t resist haha
so, i tried it with Pro Tools using slate VMR(oversampling on every instance) with a 24 track session, it nulls perfectly, both with 2 offline bounces, 2 online, 1 off 1 on, every combination.
the only way i got something out of it was by using Plugin Alliance SSL channels and randomizing the tmt channels between bounces(duh, of course, it changes channels on the console).
the only way i managed to get something out with the null test between two bounces was by applying reverb on the snare and vocals with lustrous plate by slate. little plate by soundtoys cancelled out perfectly
Fascinating.
Very interesting, perfectly sensible. I’ve done this test on my own, but not with modeling plug-ins. It’s the exact same process, bit for bit, either way - of course.
Always thought the real-time bouncing was meant if you’re using Hardware... then some guy told me it’s better quality... either way always bounced offline since I’m working ITB only 🤣🤣🤣
Some plugins have different real-time vs offline oversampling behaviors. For instance, Soothe2 by Oeksound has a feature where you can set different oversampling settings for playback vs bouncing. So you may be using a slightly lower quality processing for playback when mixing, so that you save processor, but then when you go to bounce the really high quality processing occurs. that's at least something to consider
Decent video!! 🤟🏾
I would imagine this is as long as ones latency is really working & on-point. As I've found this can cause subtle glitches in real-time bounces (being more detectable in low frequency tracks) if the buffer size is on the edge of what it's happy with.
Also modulation FX, operated by free runninf lfos, can cause difference between bounces (online or offline). So Phasers, Choruses, modulated Reverbs, Modulated Delays etc... can bring to a null test fail.
I think slate not “nulling” is also due to the random noise floor they add to their plugins (i think for anti aliasing reasons).
Program dependecy on a compressor can also bring digital compressors to fail the null
With Superior Drummer and similar drum softwares, they are randomizing the drum hits for realism and therefor often the bounces to NOT have the exact same samples played when you bounce them.... Just something some people might not have noticed
It obviously doesnt matter for the production, but in a phase reversing null test, this WILL show.
The more obvious thing is chorus/flanger/modulation effects. I did a null test between two live bounces; I hadn't touched a thing. The latency of my hardware reverb was off a few samples, and all the chorus effects meant the resulting null wasn't much of a null at all really.
Yes, a lot of modulation based effects don’t always start at the same point in time, so I guess in that case it might be worth auditioning more than one bounce?
@@PresentDayProduction Sort of like a synth LFO either free running, or retriggered? so that on a bounce it could hit different points in the phase of the modulation wave source? Sounds plausible.
Fantastic discussion. I found no difference in Pro Tools when I did my listening tests, though I still usually bounce real-time to catch problems last-minute.
Absolutely, real-time is great for a final listen. We do the same!
I usualy bounce in real time, not because it could be better but because it gives me the possibilty to hear "one more and a last time" what I just did.
And sometimes, well, I trash the bouce and go back to work on a final detail I haven't heard before.
Yep, using it as a final listen-through is something we do too!
Yup! Same here!
Also, it's usually going to be a "What You Hear is What You Get" scenario, which gives me that little peace of mind that I didn't mess anything up. There are few things more embarrassing than sending a out a mix or a master of glorious silence cause you had routed something wrong, or chosen the wrong bus; or had accidentally unmuted the click track or something.
Yeah that’s the main reason I do listen-throughs. To check for mistakes I may have missed, things I haven’t fixed, or incorrect routings. Likewise, I do the same in every video render I do. I watch through it on EVERY render to check for messed up frames, pixels, odd colouring, noise, etc. Rarely is there an issue but I like the peace of mind.
@@PresentDayProduction I do the same with my videos, not in real time as iMovie* doesn't show th real time rendering, but as soon the export is done I view it few times.
*(I'm not a video maker, as you can see on my YT channel my videos, are simple visual illustrations for my music only)
Hey guys just hitting you up to say if the Internet was invented when I was starting out your channel would have helped me no end. Excellent sensible resource for young kids starting out. Love your myth busting stuff - especially about plugins. Great stuff guys. Indecently, your video about mic pres will save those starting out from spunking their money on useless shit for a project studio, cos some Sound Cloud ‘ influencer ‘ ( clown ) hyped it . Great records are simply great ideas frozen in time. Can’t polish a turd. There is one exception with mic pre’s and that’s the boutique end of the market when you’re buying a very colorful one on purpose. I use a pair of Shadowhills through which I might pass stereo keyboard tracks to give em a bit of mojo. But they are very scenario. specific. Sincerely. P de H
I think you could have frozen the tracks before bouncing to make sure the plug in performance is identical
Thanks
Bravo chaps
Quality Video
Brilliant work busting this myth.
I can hear a big difference in the low end. The realtime has more punch and detailed lows. Nice video 👌🏻
This was really well done and explained - you guys have the confidence to be wrong that comes from being very good at what you do.
I posted a link at the Reaper forum and got some interesting responses including one which mentioned other factors affecting on and off line renders.
Eg some plugins actually work differently if they know that it’s an offline render (eg oversampling) or some other calculations, many plugins have different quality modes (quicker for real time so to not tax the cpu , slower but better for off line) ; some DAWs allow the user to set a different buffer/block size for offline renders which may lead to slight differences arising from automation settings etc. There is also dither noise if employed which would not null.
That make sense?
Just tried it with Reaper. I did the mix almost completely with Waves Analog simulated plugins (SSL strips, J37 tape, Pultec EQ, etc) and Amplitube on guitars.
Soooooo much variation between two offline bounces rendered back to back! Super interesting.
Also I had a guitar using chorus effects through Amplitube with a rate that was not synced with the tempo. That guitar wasn't cancelled out whatsoever. I heard choppy bits of the whole song like in your example, but that single guitar loud and clear.
going to try this with Pro Tools, just for fun, both with Slate's and with only Fabfilter's plugins, i don't expect the results to be different but we'll see, or hear...
i usually bounce online(track recording inside the DAW) because my starting template is structured so that i print stems, instrumentals, acapella, and TV track in a single pass.
give the cat a boop from me.
🎵 Scientists in a lab, making up a logo, stop doing that and go and make some more vaccines 🎵
I will be singing this for DAYS.
This creates a conundrum. Do certain plug-ins sound different on each playback...? "...I'm sure it sounded better 10 minutes ago...I'll never finish anything ever again" But no, wait, I use Logic X stock plug-ins...Pfff, I'll still never finish anything ever again...
You seem to get straight to the point with no messing on these things. I need you ! Keep up the myth-busting.
If you feel you can't finish a mix, try to start from scratch, excepts for the edits you've done.
In many instances it has helped me, and after a few times I realized there were a few things I was overdoing or approaching the wrong way and I stopped doing those mistakes.
@@kelainefes , thanks for your tips...this was just my poor effort at humour... 😉
@@thelonetwangster Hahaa ok sorry, I thought you were joking about the classic "I can't finish the mix" issue but also being half serious.
Glad that you're past that phase mate.
Evidently I wasn’t wrong with thinking offline is better because it’s faster. Thank you so much for the confirmation!
Yes. I have reached the same concluído. I am a experienced studio owner producer.
I just did this with pro tools and I got 100% silence with phase reversed, meaning they are exactly the same.
How does this work?
@@mazhet6572 If two signals are exactly the same when the polarity/phase is reversed on one of them they will completely cancel each other out resulting in complete silence,.
@@archtop thank you
Did you use any analog emulation plugins like they did in the video?