Elastic audio revisited and reviewed
So 7.4 and elastic audio have been available for some time now. As you might know, my initial feelings about it was great. After several days of working in it and with it my opinion is unchanged.
Elastic audio and loops
You can spend a lot of time making loops fit together. For instance, I had just recorded 12 bars of bass guitar. As often is the case when I’m playing, it had a lot of syncopation and not your typical straight 4/4 feel. I imported a jazzy electric piano loop at random. Now, I figured they’d be pretty close in style with the loop having the word “jazzy” in the name and all. No way. Very different feel. It was also played around 35 BPM too slow. So how do you practically solve this?
1) Enable elastic audio on the track. For this particular file I used the polyphonic algorithm.
2) Take the stretcher tool and speed it up so that it fits the grid. I actually managed to do it the other way around as well, i.e. making it slower and thus twice as long as the first stretched result. Both sounded good! Amazing. The even easier way is to simply right-click the region and choose “conform to tempo”.
3) Cool, they’re locked to the same tempo. Unfortunately my playing was a lot more jazzy than the jazzy loop. Inserting markers and moving them manually is actually very quick once you get used to it but the real money is in the automatic way. Option/alt+3 (numeric keyboard only) is your new friend. Enter the quantize window!
Elastic audio and your typical rock band takes
So how well does this type of stretching work in a typical editing situation with real instruments. I’ve tried it on pretty much everything and there are both wo-ho:s and no-no:s.
Percussion: Wo-ho! I’ve seen some people complianing that it sounds bad on drums. I don’t know what kind of stunts they’re pulling. Works really good here.
Bass: Wo-ho! As expected, it works great on bass.
Acoustic guitar: Wo-no. Really depends on the type of playing.
DI:ed or clean recorded electric guitar: Wo-no. Worked better for me. Probably depends on your playing too.
Slide guitar: No-no! Don’t even think about it.
Keys: Wo-no. Tough call, depends, I guess.
Vocals: Wo-no. I had success with it. Other haven’t. As you know, the human voice is a tough one, pitch and style variations are huge. From semi-spoken to Mariah Carey wailing. If the take sucks then try to stretch it a little, it works sometimes, other times it probably won’t.
Is this a new age?
Yes. I really think so. Time-stretching have been available for some time in all kind of forms, but I really believe the Pro Tools way will change a lot. I know my workflow has changed, and I know a lot of other peoples workflow have changed. I think we in time will see something similar in other DAWs as well. Elastic is the future, and I can’t wait to see elastic pitch.
I really like Elastic Audio as a tool. The problem is when you reach for it rather than trying to get a better take.
By Jon ~ December 31, 2007 kl 1:12 pm
I’ve edited and re-timed vocals, drums, even complete orchestras. In my experience, the hardest things to re-time are instruments like guitars and harps when they’re playing broken chords and arpeggios. I’ve had 7.4 for a month, but I’ve been busy mixing, rather than editing. How good is it on this kind of especially problematic material?
By Guy Sigsworth ~ December 31, 2007 kl 3:49 pm
Guy: I’ve been using it on an EP that I’ve been working on the last couple of weeks, and it actually works well. In the cases where say, the guitarist exaggerates a strum on the guitar, well you can either drop in markers for each transient of each string hit, or what I’ve been doing is just dropping in a marker at the last string hit, and using that as the first beat. you’ve got lots of control, actually.
I will say this though, the most important thing that I would stress about Elastic Audio is the same thing that’s been true all these years about other time/pitch plugins like Serato Pitch N Time: Garbage in -> Garbage Out. If you’re working with 16/44 sessions, the longer you timestretch something, the worse it’s going to sound. I’ve been working 24/48 sessions, and they sound really good. I’ve played around with some 24/96 files and they sound really good too. You’ve got to have good source material.
I don’t know exactly how elastic audio works, but if it works anything like the graphic world, when you resize say, a JPG file, you either throw out pixels when you downsize, or (depending on the algorithm) create new pixels when you enlarge. I’m guessing that in time stretching, it’s doing something similar, where it’s looking at one sample, then the next sample, and creating a new sample in between to approximate it. If you listen to something that’s been really stretched in elastic audio, it sounds alot like granular synthesis, like it’s being processed in Reaktor. I think that’s why they include controls for the different time stretching algorithms; to control the granularity to make it sound more natural.
I still think using Serato Pitch N Time sounds better because it processes it offline as an Audiosuite effect, but it takes alot of time. I’ve retimed whole multitrack drum sessions a beat at a time with Serato, and it takes days. With Elastic Audio, it takes hours.
I wouldn’t want people to think that this is a magical tool where you can simply select all of your regions and quantize it all and you’ve got instant success. You have to have good source audio, good incoming transients, and really you should have a good performance to start out with, and even then, you have to make some tweaks. It’s not perfect at detecting transients, and you may have to customize it a little. But it’s definitely worth the upgrade, and it’s caused me to rethink the way that I track and set up sessions (to be able to take the most advantage of elastic audio) and I think that makes it a pretty noticeable feature.
By Chris Hahn ~ January 1, 2008 kl 2:10 am
Excellent comment Chris!
Guy, if I remember, I’ll try to stretch some arpeggios the next time I fire up Pro Tools.
By stiff ~ January 1, 2008 kl 8:29 am
Chris
Thank you for that thoughtful ad detailed reply. It’s a really exciting development of Pro Tools. Some of what 7.4 is doing seems a bit like Melodyne - some kind of granular resynthesis thingy, perhaps? The next stage must be a pitch detection algorhythm.
As you say, the existing time stretchers can sound excellent if the source material is well-recorded. I own every time stretch plug-in out there - Serato, Melodyne, Waves, Speed and the Digidesign plug-ins. I choose differently depending on the task. If I have to speed up a vocal, say for a remix, I usually edit back in, non-time-stretched, any long notes with vibrato. Speeded-up vibrato can sound unintentionally comedic, like a kind of cyborg Edith Piaf, and I usually don’t want to subject the singers I work with to that feeling of caricature.
By Guy Sigsworth ~ January 1, 2008 kl 7:17 pm
Greetings,
Just found this blog from the DUC. Very cool stuff Thanks!
Chris, Your post was really helpful and informative.
I have been with PT since 4.3, have just installed 7.4, but I have yet to explore the function or implementation of EA. I was hoping you might have some advice on how you are setting up sessions and tracking with EA in mind.
In my work, I will probably not use EA for any substansial tempo changing, but more for tightening up timing and nudging things around.
I do some work with full band instrumentation(drums, egtr, keys, etc) but most of what I do are acoustic sessions(guitar, mandolin, violin, etc) so there is no room to hide anything. Any artifacts in this process will be clear to all ears.
I am a little hesitant to jump in to EA on the clients time until I am confident both I and EA are up to the task.
Thanks again, Cheers to all!
By JP ~ January 2, 2008 kl 1:26 pm