Crappy elastic audio sound quality?

[AlexLakis]

I ran into a snag. I gave the M-Audio FW410 I use on my PC M-Powered 7.4 rig to a friend last month, and I assumed I could use my M-Audio FW610 with it as needed, but it turns out the FW 610 is NOT SUPPORTED by 7.4! Doh!

I ordered a M-Audio Firewire Solo on Ebay, and it should be here by Aug. 5th. I plan on posting the test files as soon as I receive it!

Please stay tuned, and thanks again for your patience!

[stakeoutstudios]

I absolutely agree that elastic audio worked and was usable when it came out in 7.4 - since 8.0 it has been pretty much useless. I do wish Avid would fix it, but as it's been broken so long, I don't hold out much hope.

[AlexLakis]

Thank you all for your input...Its good to know I'm not the only one who is concerned with this!

The Firewire Solo has arrived! I should be able to complete the X-Form tests later today! I can host the .WAV files on my server, unless someone has a better idea.

[AlexLakis]

Here are the files:

http://Clients.KillerPenguinProductions.com
login: DUC
pass: Quackquack1

Password is case-sensitive! Files are 44.1/24 .WAVs.

I included the original track; a recording of me playing a scale on electric bass guitar (a little bit sloppily.) Its a Mexican Jazz Bass into the DI on my M-Audio ProFire610.

The subsequent bounces are files edited with Elastic Audio, and committed using X-Form. These files are processed using the EXACT SAME methodology; the only difference was one was done in Pro Tools 7.4, the other in Pro Tools 12.

In hindsight, I should've played the bass even more sloppily than I did, but the differences are still apparent, even with the minor quantization adjustments. Notice the softness of the attack in the PT12 bounce, as opposed to the PT7 bounce, which retains the transients much more accurately. The 12 bounce also sounds a little more "saturated" to me. These differences are apparent in other applications too (multi-track drums), but you should be able to clearly hear the difference. My wife just did a blink ABX test with me where she played one of the two X-Form'd files randomly 10 times while I faced away from her, in headphones (Sennheiser HD280 Pros); I was able to tell her which file was which 10/10 times (100%).

Alexey (and everyone else,) please let me know what you think! Like I said, I can make the difference sound more pronounced if you can't listen to the files in a critical environment.

Also, Alexey, I thought about your potential solution about moving the iZotope_Radius_*.* binaries from 7.4 to 12, but wouldn't they be *.dpm and need to be *.aax?

Thanks for indulging a concerned user!

[mesaone]

Originally Posted by AlexLakis

Also, Alexey, I thought about your potential solution about moving the iZotope_Radius_*.* binaries from 7.4 to 12, but wouldn't they be *.dpm and need to be *.aax?

The code path (32 bit vs 64 bit) is a problem. From what I can see, the Xform AAX is just a hook to the 64-bit VST plug-in located inside the AAX itself.

C:\Program Files\Avid\Pro Tools\System Plug-Ins\Elastic Audio.aaxplugin\Contents\x64\iZotope_Radius_XForm_x64.dll

And if I recall correctly, the 32 bit radius DLL file for PT10 and under was located outside of the plug-ins directory. So transplanting from 7.4 to 10 might work, but 7.4 to 11+ would not.

[Alexey Lukin]

Thanks, Alex. I do see transient smearing in the latest result. What was the processing that you've applied? I see that the resulting files slightly differ in length: one is shorter than the original, while another one is longer. Timing of some transients also differs.

[AlexLakis]

Thanks for the reply, Alexey!

Not sure why the resulting files differ in length. I just selected the waveform block and bounced. Most likely, the Elastic Audio operations shifted the end of the block.

My methodology for single tracks like this is as follows:

1. Make sure tracks are in "ticks" mode.
2. Enable elastic audio. Monophonic in this case.
3. Using the default settings, analyze and generate warp markers.
4. Select the waveform block and quantize elastic audio events.
5. Select X-Form in the EA menu to go into offline mode.
6. Select Remove and Commit.

And, to reiterate, the same exact workflow was followed in 7.4 as was in 12.

Please let me know if I'm "doing it wrong," but I've been using X-Form successfully in 7.4 for years now doing exactly what I just outlined with fantastic results (sometimes multi-track drums require an additional step or two and/or manual warp marker generation). The results just aren't there for me in 8-12.

[Alexey Lukin]

Alex, does the quality difference between PT 7.4 and 12 exist with simpler scenarios? Your current workflow involves markers and quantization, and I see that the results are different, timing-wise. Can you try some simpler operation, like a fixed-ratio time stretch by a certain percentage on this sample, without any markers? I'm wondering if the quality difference is related to the fact that you are using markers.

[AlexLakis]

Sure, I can post examples of a fixed-ratio time stretch tomorrow!

I rarely work with Elastic Audio/X-Form without warp markers. It just makes me wonder why the same workflow produces such different results, aside from the algorithm itself being changed between PT 7.4 and PT 8+?

Thanks for hangin' with me!

[dieffe]

my two cents: I almost always use elastic audio with the warp markers too; that's the real strength of it.
I can't test PT 12HD against other, older, releases, but I definitely can hear the processing (especially on dialogue, and I work with dialogue most of the time)