Audition 1.5 - (Much more than a) Centre Chanel Extractor

[tenebrenz]

Adobe Audition 1.5 has a new feature called "Centre Chanel Extractor" which allows the extraction or removal of sounds based upon where they appear in the stereo field. My intial tests on audio have found it very good at this, for example on a recording with piano left, vocals centre and guitar right it is possible to extract these into individual waves. You could then remix them back together however you like. The tool also allows the extraction of surround audio from a matrix mix.

Though I don't have a 5.1 system, or space for one, - I think this would be worth investigating if you're into Stereo/Surround to 5.1 conversion.

[ursamtl]

Originally posted by tenebrenz
Adobe Audition 1.5 has a new feature called "Centre Chanel Extractor" which allows the extraction or removal of sounds based upon where they appear in the stereo field. My intial tests on audio have found it very good at this, for example on a recording with piano left, vocals centre and guitar right it is possible to extract these into individual waves. You could then remix them back together however you like. The tool also allows the extraction of surround audio from a matrix mix.

Though I don't have a 5.1 system, or space for one, - I think this would be worth investigating if you're into Stereo/Surround to 5.1 conversion.

Now this sounds quite promising! How about the sound quality of the remaining instruments after you extract one, is it still good? I ask this because most "vocal zappers" I've heard are effective at removing vocals but they do so by basically just substituting a stereo difference signal. Depending on the source material, these may consist mostly of ambience or instruments that are hard panned to one channel or another. However, if Adobe has figured out a way to extract the sound at certain positions without ruining the rest, then that's very, very cool!

[tenebrenz]

It uses FFT phase analysis and other stuff I don't understand. It can give artifacts (similar to excessive noise-reduction) if you try to isolate a sound too much so you have to use some common sense, but it really is impressive. For example, you can remove/reduce vocals from the centre (or anywhere else) and keep everything not at this point in their original stereo placement.

[ursamtl]

Originally posted by tenebrenz
It uses FFT phase analysis and other stuff I don't understand. It can give artifacts (similar to excessive noise-reduction) if you try to isolate a sound too much so you have to use some common sense, but it really is impressive. For example, you can remove/reduce vocals from the centre (or anywhere else) and keep everything not at this point in their original stereo placement.

How is the bass response? This is often the first thing to suffer.

[tenebrenz]

Well, if bass is present at the same stereo position of the sound you wish to extract / remove it will be processed. You can, however, restrict processing to a specific frequency range and leave the bass intact - the standard presets do this by default.

I'm not a rep for Adobe, just thought you guys might like to try this out!

[ursamtl]

Originally posted by tenebrenz
Well, if bass is present at the same stereo position of the sound you wish to extract / remove it will be processed. You can, however, restrict processing to a specific frequency range and leave the bass intact - the standard presets do this by default.

I'm not a rep for Adobe, just thought you guys might like to try this out!

Sounds really good. I'll have to try it out. And no, you don't sound at all like an Adobe rep.

[PeterC]

Been fiddlig with the center-channel extractor, really fab and effective. However doing a stereo track to surround with its aid is a pain to me, since my pc has 2 speakers (destination is a 5.0 system). I just can't find the ideal levels for each channel. But I must say the latest Audition is great.

[ursamtl]

Originally posted by PeterC
Been fiddlig with the center-channel extractor, really fab and effective. However doing a stereo track to surround with its aid is a pain to me, since my pc has 2 speakers (destination is a 5.0 system). I just can't find the ideal levels for each channel. But I must say the latest Audition is great.

You might try taking DVD with a surround mix you like and ripping the sound from it and then using some sort of meter to measure the sound level for each channel. I know it's a pain without being able to monitor, but with a bit of practice, you can get the hang of it. I've been doing a lot of 5.1 mixing on my PC, which has only 4-channels and I'm getting really accurate mixes. Depending on your preferences, the surrounds are usually not as loud as the fronts.

Let me know if this helps and if not, keep asking!

Regards,
Steve.

[PeterC]

Thanks for the tip. Tried it out, but I guess either I'm just fiddling with the wrong pieces. I get way too much surround activity, although on the song I've been testing with, I like certain "effects" it brings out.

I took levels of Queen - '39 of the 'A Night At The Opera' dvd-audio and I got something along Center being -3 peak, front -3.5 peak and -6 peak for surrounds. That didn't really work for my elements (well duh)

Also wrote a PLII file with the nice guide here and took levels of -12.5 -9 and -3 (S, F and C). Those also resulted in way too much emphasis on the surrounds and I use a default -3dB attenuation on my surrounds in the amp.

Perhaps I need to re-do my elements. The song in question is "Already Gone" by Eagles, I've separated the center, which is mainly lead vocal, chorus, guitar, drums. The fronts I use are mainly just "vocal removed" and contain drums, guitar etc, and the backing vocal (ooooooooo part of the song) and the surrounds I got are the most extreme left and right portions of the fronts (mainly guitars, very little drums). What I enjoy about this is the drums on the intro go around the room from SL to SR to FR (and so on) on this nifty mix I've done, but I guess I've tried to do too much separation between the channels. The rhytm guitar from the left and the hihat from the right become way too dominant for my ears, however lowering them breaks up the whole thing. So, maybe just maybe I'm overdoing this ;-) And should do less "discreete" channels.

Has anyone done songs with this, a good guide would be nice
:D

[ursamtl]

Hi Peter,

I know what you mean about liking the effects the surround channels can bring out. Back when I was a kid in the 70s (there I've dated myself now :), I discovered this circuit in an electronic magazine that gave a cool surroun-like effect by taking a speaker and hooking it up to the two positive terminals of a stereo amp. This gave the "difference" signal between the two channels, which is often ambience. It's also often brings out things like guitars and background vocals much better than the regular stereo mix. In fact, the original Dolby Surround was based on this. So now when you play with this in Audition or whatever software you're using, the same sort of thing occurs, you find things sort of "hidden" in the music that you might not have realized was there.

As for the levels, just keep trying until you get something you like. Always remember that each song is different, so the results are going to be different. For example, the early Eagles stuff was generally recorded fairly "dry" especially on the drums, so there's not a lot of ambience to play with, but the mixes are good so there are lots of hidden nuances in the instruments and of course those great harmony vocals! I'm wondering how you "separated" the different elements. Maybe part of the problem is when you put them back together in the new form. Sometimes what sounds great individually interacts with other sonic information and through phase and frequency interaction, cancelling, etc., the desired combination just doesn't happen. By the way, I also bought the Queen Opera DVD-A and I must say I was quite disappointed with it. At the same time I bought the DVD-a of Pet Sounds and was absolutely amazed at how well they took what was originally mono and turned it into really beautiful, full surround sound.

As for a guide, there are actually a few on here. You mentioned the DPLII one. There's of course the Ambisonics one but peraonally I didn't have good results with that. It gave a surround effect but it sounded something like taking the entire stereo mix, making it mono and then recording the mono recording playing in a room with very little in the way of reflective surfaces. But the again others like it. Personally, the best sound I've heard so far is from something kempfand put together based on theories put forth by a famous audio expert, Michael Gerzon. There is not yet a guide for this but if you follow the original ambisonics guide to learn how to use the Plogue bidule software and then download the Gerzon_1997 files, I think you'll like the results. I'm actually working on something now that is sort of a hybrid of this Gerzon approach with something else. Once I put it all together, I'll write a guide.

Have fun!
Steve.


Originally posted by PeterC
Thanks for the tip. Tried it out, but I guess either I'm just fiddling with the wrong pieces. I get way too much surround activity, although on the song I've been testing with, I like certain "effects" it brings out.

I took levels of Queen - '39 of the 'A Night At The Opera' dvd-audio and I got something along Center being -3 peak, front -3.5 peak and -6 peak for surrounds. That didn't really work for my elements (well duh)

Also wrote a PLII file with the nice guide here and took levels of -12.5 -9 and -3 (S, F and C). Those also resulted in way too much emphasis on the surrounds and I use a default -3dB attenuation on my surrounds in the amp.

Perhaps I need to re-do my elements. The song in question is "Already Gone" by Eagles, I've separated the center, which is mainly lead vocal, chorus, guitar, drums. The fronts I use are mainly just "vocal removed" and contain drums, guitar etc, and the backing vocal (ooooooooo part of the song) and the surrounds I got are the most extreme left and right portions of the fronts (mainly guitars, very little drums). What I enjoy about this is the drums on the intro go around the room from SL to SR to FR (and so on) on this nifty mix I've done, but I guess I've tried to do too much separation between the channels. The rhytm guitar from the left and the hihat from the right become way too dominant for my ears, however lowering them breaks up the whole thing. So, maybe just maybe I'm overdoing this ;-) And should do less "discreete" channels.

Has anyone done songs with this, a good guide would be nice
:D

[PeterC]

I've done the separations solely from a 32-bit 2-channel wav with the center channel extractor, tweaking from the presets (such as acapella, vocal cut) to obtain results that don't contain artifacts (or anything noticeable). I think my main fault with the song is simply as I mentioned too much separation attempted, they don't blend, it hardly sounds like a song, but 5 elements arguing.

I've tried the bidules here, but it hasn't really been that good on what I listen. The SAD5.1 was nice, but I'd still prefer Pro Logic II over these (maybe I'm too used to it, I like centered vocals in general)

offtopic: Disappointed with ANATO dvd-A? Wow, don't hear that often ;-) I know the 1st US release of the DVD-A got poor reviews and as I recall it wasn't supervised by Brian May or Roy Thomas Baker (might be wrong, the second was revised as I read on some fan page). The second version however was raved about and I personally think both Queen dvd-audios are everything their cd releases mainly are not. Simply great sounding. :D

[Sycho]

has any one tryed inverting the right channel and doing the center channel extracter to remove a surround channel

[ursamtl]

Originally posted by sycho
has any one tryed inverting the right channel and doing the center channel extracter to remove a surround channel when you say "remove" do you mean to extract sonic info to create a surround channel?

[ursamtl]

Originally posted by PeterC
I've done the separations solely from a 32-bit 2-channel wav with the center channel extractor, tweaking from the presets (such as acapella, vocal cut) to obtain results that don't contain artifacts (or anything noticeable). I think my main fault with the song is simply as I mentioned too much separation attempted, they don't blend, it hardly sounds like a song, but 5 elements arguing.

I've tried the bidules here, but it hasn't really been that good on what I listen. The SAD5.1 was nice, but I'd still prefer Pro Logic II over these (maybe I'm too used to it, I like centered vocals in general)

offtopic: Disappointed with ANATO dvd-A? Wow, don't hear that often ;-) I know the 1st US release of the DVD-A got poor reviews and as I recall it wasn't supervised by Brian May or Roy Thomas Baker (might be wrong, the second was revised as I read on some fan page). The second version however was raved about and I personally think both Queen dvd-audios are everything their cd releases mainly are not. Simply great sounding. :D

Yes, it sounds like maybe you did too much separation. As for the bidules, different people get different results. Some that others have said were great sounded terrible to me, and vice versa. Actually, DPLII is not bad at all. I've corresponded with an audio engineer who rated DPLII a 3/5 in comparison to 1/5 for processing regular stereo files with Ambisonic encoding/decoding.

My problem with my copy of ANATO DVD-A was the sound quality, especially high frequencies. I found actually switching to 2-channel sounded cleaner. I remember reading about the difficulty they had finding the original tapes for all of this album when they did the remix, but overall I'm not as impressed with the sound as I thought I would be. I know wonders can be worked with old material, witness the amazing job Kevin Shirley did with the Led Zeppelin DVD and How the West Was Won releases last year. I was just disappointed with ANATO. Perhaps I have one of the original releases.

[Sycho]

Originally posted by ursamtl
when you say "remove" do you mean to extract sonic info to create a surround channel? yup

[PeterC]

Originally posted by ursamtl
My problem with my copy of ANATO DVD-A was the sound quality, especially high frequencies. I found actually switching to 2-channel sounded cleaner. I remember reading about the difficulty they had finding the original tapes for all of this album when they did the remix, but overall I'm not as impressed with the sound as I thought I would be. I know wonders can be worked with old material, witness the amazing job Kevin Shirley did with the Led Zeppelin DVD and How the West Was Won releases last year. I was just disappointed with ANATO. Perhaps I have one of the original releases.

On this offtopic issue ;-) If your ANATO DVD-A copy was/is

Ordering number (print):
69286-01091-9-3 DTS-1091
Ordering number (matrix):
IFPI LL07 B01K1635 DC-1091 03
IFPI LL07 B01K2028 DC-1091 23

You have the 1st release. Also I think the 1st release
didn't feature the Bohemian Rhapsody video as an extra.

[PeterC]

Alright, here's what I've done with the Center Channel Extractor for Eagles - Already Gone, that actually sounds somewhat tolerable in my setup.

I use the VOCAL REMOVE preset to R and L at -5dB or so.
These are FL and FR.

For center channel, LIFT VOCALS 10dB
preset.

Surrounds I use, DISCRIMINATION=0, Center Channel Level -12DB, Full Spectrum. On all cases I have the Spectral Decay Rate at 98% + some other minor settings (everyone can tweak them to their own ears)

Then put them all in the Multi-channel encoder,
set FL & FR to their own spots (mono-files), Sl & SR too
, C too, then I have a SECOND CENTER file which I pan to FRONTS and Surrounds to slightly increase the vocal presence there.

Then I set the levels of the WAV files exported
to FL & FR normalized to -7dB, CEnter to -3dB and Surrounds to -12dB (actually 0db c,fl,fr and -3db sur seems much full). Ready to encode to DD or DTS. (NOTE: adding some reverb to the surrounds in the case of this song was rather pleasant)

Note: I have no possibility of real-time monitoring, those who have might come up with "brighter" values for levels etc.