A few days ago I read a great post on “worldizing” by Andrew Quinn. To be honest, I hadn’t heard that term before, but I understood the concept. Worldizing is an old technique used to replicate the quality of a sound played in a space or piece of equipment. To do this, the original recording is played back in an environment, and the result is recorded.
Today, convoluted reverb plugins like Altiverb, Waves IR, TL space, etc can achieve the same result with much less work. With these plugins, an environment only needs to be sampled once with an impulse response, and it can be used on anything you’d like. Andrew’s demo made me wonder how true convolution reverb is to worldizing. Is there a noticeable difference between the two?
To test this out, I decided to try out and compare both techniques on a few things. To get an idea of the reproduction of different amounts of reverb, I worldized/processed my bathroom, living room, and stairwell. I also ran tests on a cell phone, TV, and pillow speaker. In each test, the sound source and recorder remained in the exact same relative position for worldizing and impulse response. You can check out the results below. Each example has alternating, worldized, and processed versions.
Note: I used voice-over lines in the rooms, and music for the devices. I made this choice because I was primarily concerned with decay time for rooms, and frequency response for devices.
The processed version sounds pretty convincing, however the worldized version has a noticeably longer decay. I want to note that I made a complete new IR for this stairwell even though I had one previously made. Again, I wanted to make sure the sound source and recorder were in the same position for both tests.
I have a very hard time picking out any differences between these two. Decay time and frequency response seem almost perfect to me. If anything, there’s a touch more high-end in the processed version.
The bathroom also turned out really well. Decay time in the processed version was spot on again, however the worldized recording had less high end when compared. It’s easier to pick this out on the alternating version.
I’m pretty satisfied with this one. Both the worldized and processed versions sounded very similar. The one difference I noticed was the processed version having a little more low end than it should. You might notice this in the third and sixth changes in the alternating version.
Another convincing performance. However, the processed version lacked in the hints of inherent distortion of the cell phone’s speaker.
If I had to pick the least comparable example, this would be it. Listening to the two separate, it might not be apparent. However, there are some noticeable changes in frequency response between the switches in the alternating version. The worldized version has more energy in the low end and the processed has more in the high freqencies.
I was pleasantly surprised with the overall performance of convolution reverb. At the same time, worldizing was a blast to try out. Nothing beats freaking out neighbors by blasting GlaDOS lines in an apartment stairwell.
There was only one instance of noticeably different decay, and that was the stairwell. This one really surprised me. I can’t definitively blame this on the program, it could be human or equipment error. Perhaps level differences between the sine tone sweep and voice over led to differing results.
When there were discrepancies in frequency response, the processed version tended had more high-end. Now, there are a number of reasons this might happen. In the case of rooms, it’s possible that my playback equipment wasn’t quite up to par. However, I don’t understand the frequency differences when it comes to the devices. The devices don’t play back with a flat response, but that’s exactly what the plugin needs to replicate it.
One thing I new convolution reverb couldn’t replicate was distortion. I knew this before hand, so I tried to keep all of the devices’ volume at an appropriate level., but it was still evident in my cell phone examples.
When it comes to applying these techniques, it’s a question of preference, time, and money. Maybe the authentic distortion of that particular cell phone is important to you. Many times you may want the inherent “flaws” (I use that word lightly) of a space or device. While it’s certainly possible to add distortion and decay to convolution reverb, you might not have known to do this if you hadn’t heard the worldized version in the first place. However, worldizing does take a lot of time. It needs to be done for each and every piece of audio you want to use. Maybe it’s not worth the time to do all that work for simple rooms. Convolution reverb offers a much quicker turn out time, and (this is a big one) allows for changes to be made to the dry recording. It’s much harder to edit a worldized recording. If changes need to be made, you might have to worldize the whole thing again.
I really recommend checking out Andrew’s post on the subject. He has some great videos and links to even more information.
Sine tone sweeps
Bathroom: 30 second sweep + 3 second silence
Living Room: 30 second sweep + 3 second silence
Stairwell: 30 second sweep + 7 second silence
Cellphone: 30 second sweep + 3 second silence
TV: 30 second sweep + 3 second silence
Pillow Speaker: 30 second sweep + 3 second silence
Playback (Rooms): M-Audio AV40
Recording: Zoom H4n (48kHz/24bit)
Processing: Audio Ease Altiverb 6
Cellphone: Motorola Droid 2
TV: Panasonic TC-P42G10
Pillow Speaker: Unknown