Everyone knows Altiverb is great. I've tried it - sounded absolutely gorgeous on vocals where some algo reverbs on the same source seemed to do things with the tail that sounded somewhat "fake" to me. However, what are the criteria for comparing one convo to another? Is it just the impulses that make the difference? Seems like that would make the $600-900 price tag of Altiverb pretty ridiculous. When I use the convolution reverbs built into Cubase and Live they sound better than most algos but yet not as good as Altiverb.

...why?

@Von_Steyr I feel like this is the sort of thing you would know about. Or maybe @Baxter @superliquidsunshine or @rhythmatist ... you guys are the main reservoir of knowledge that I tap into when my relative inexperience fails me. Help a brother out?

 

Quite simple.Convolution reverbs are real, they actually went there with gear and many mics and recorded the acoustics of a room.
Algos are not real, computer code, guessing.
Both are useful, however quality vs cheap reverb can also make or break a production or make a good production amazing.
I think altiverb is worth the money.

 

Thanks for the vote of confidence, but I don't know. What Von-Steyr said seems logical. I generally only use convos for the wilder special effects. The tunnels, hallways, stairwells--unusual spaces things. For standard plates, halls, rooms, etc. I have my favorites like most people here. So easy to find what you want or tailor to your needs,----if you can come to a final choice with so many options.

 

A convolution reverb technically multiplies each single sample (I'm talking of the smallest unit of sound information, not "a sample" as in "a snare sound", for example) with all single samples of the impulse response. (CAUTION: This is a very simplified picture of a convolution)

With this easy picture in mind you can instantly see a few factors that influence the quality of a convolution reverb.

First of all, how is the impulse response recorded? You can for example just clap your hands (impulse) and record the response from the space. Or you can use a sine sweep over the whole audible range (mostly 20-20000 Hz) and record the response to the sweep. The former is a response to noise, which has a lot of frequencies of the audible range, but is pure randomness and doesn't guarantee all frequencies. The latter guarantees all frequencies but must be transformed first into a usable response file. Sine sweep gives the best quality and should be preferred.

Also, at which bitdepth was the response recorded. Less bits mean less information for the sampling process. 16 bit might be good for listening to CDs, but for high quality convolution reverbs 24 bit is the least.

And the inner workings of the algorithm is also important. Working with 64-bit internally it can convolve to a greater depth, before converting the signal to 32-bit using interpolation. That adds further quality to the result compared to an engine working with 32-bit internally.

And lastly the microphones used for recording the impulse responses are a very important factor. The more linear they record, and the higher the frequency range they can cover with linearity, the better the results.

In other words: If there are two convolution reverbs that work with 64-bit internally and the same interpolation algo, and you have a set of at least 24-bit impulse responses recorded with a sine sweep and top-notch mics for both of them, you won't hear any difference. But exchange just one of these preferences, and one will have a better quality than the other (let's say a 32-bit engine, or other response files for one of the engines, etc.)

 

@Von_Steyr yeah I get the difference between algorithmic and conv. verbs. what confused me was the difference from one convolution verb to another. Since they are all just superimposing the qualities of the impulse responses, I wondered if there was some x-factor beyond simply the quality of the IRs that make them better or worse

@tulamide that makes sense. The in-depth explanation is much appreciated

 

btw @rhythmatist agreed about using convolution for cool effects. One of the (many) reasons I think Izotope Trash 2 is so great is because it has a convolver

 

Different gear, different mics, different techniques, different rooms, plugin code, optimization, etc...

 

Definitely the internal resolution and the applied interpolation algorithm.

 

Marketing, overhyped advertisements, superbeautiful GUIs, words like "it is extremely analog warm and even much better" make one product better than another. Apriori.

Dsp coding, material and results, testing and analysis, tastes, comparison make one product better than another. Aposteriori.

Or something like that...

 

I think this is about right, algos still have a long way to go to in replicating the seemingly in-regularity's involved with reverb.

Having said that, as soon as I start to turn knobs on a convolution reverb the real picture that just blew me away starts to fall apart.

 

Off topic.....I remember an interview with a famous house producer, forget his name. The interviewer was going on about his analog sound and what harware and software that the producer might be using to get his sound.
He laughed and said "I use the built in eq in logic. I just boost the mids."
Mind blown

 

Last edited: Nov 20, 2016

APART FROM the IRs, I think it's the controls & the DSP. SInce those are the only other possible variables.

Tried loading the IRs from altiverb into cubase or v/v?

It's not 32 vs 64.

If you can't hear the diff btw 16 & 24 bit, you can't hear the diff btw 32 & 64bit. 32bit flp has hundreds of dBs of headroom & millionths of dBs of precision. 64bit is only a red herring for the marketing. 32 is enough.

I don't think the SR, bitdepth, and matter that much unless the SRC sucks very much. https://audiosex.pro/posts/131883/

FFT Size might be more important, since there is a time-frequency tradeoff, and yet it is also not that important in the overall scheme of things,.

Well, marketing is another factor.

===

I'll also take a guess and pull something out of my ass:

there MIGHT be a diffrence btw 0-latency & latencied convo reverbs, just like there's a dif btw IIR & FIR (LP) EQs. Phasing diffrences.

 

Last edited: Nov 19, 2016

It seems you don't program that much, do you?

I'll make it simpler, as to why a 64-bit engine has advantages over a 32-bit engine. Let's imagine a 2-bit engine. You now can calculate with 4 states. 0,1,2 and 3
The first sample arrives. It is a 1. The IR is, say, 3, 2, 1, 0
Output will be 3
Second sample arrives. It as a 2. Houston we have a problem. The output would need to be 8 (6 for sample 2 and 2 for sample 1, which still reverberates), but unfortunately 3 is the maximum. So, output will be a false 3.

Now the same with an engine calculating in double the bit depth, a 4-bit engine. Now you have 16 states, 0-15. Obviously the second sample will be outputted correctly. Even more, it takes way longer until the bit depth isn't sufficient anymore.

As a part time programmer I find it rather funny to argument with red herring. There is not difference between 2 to 4 bit, 8 to 16 bit or 32 to 64 bit. You have double the "bandwidth". What you do with it, that's another point. But of course it matters! It's about math here, not about listening to a song.

I never talked about SR, I talked about bit depth, and again this is for math/calculating reasons.

The FFT window defines the length of the IR (or the block length, if continous blocks of FFT windows are used) that can be used. It is important in that regard. It has no effect on sound quality in convolution engines. That's different to direct manipulation effect based on FFT. Those are dependent on the window size regarding the quality.

 

Almost no difference but how an IR being processed. The plugin engine is almost same for most of the plugins conceptually. Of all the plugins I've used so far, just like @Von_Steyr said, it's down to how clean an impulse has been recorded. So again back to how well the IR was captured. Altiverb has multiple benefits apart from having very clean IRs. One of them is the positioning of source signal which is not found in many plugins apart from Reverberate, Hofa IQ, Origami and may be one or two more. However, EWQL Spaces sometimes and mostly many times takes precedence over all these including Altiverb due to extremely clean IRs despite lacking in fancy functions, one of which is early reflection. There is one more plugin which has given a concept (as far as I know) of using convolution reverb's ER and algorithmic reverb's tail. It is Vienna Symphonic MIR pro which also has positioning function. There was another one called SIR convolution reverb which was famous of having clean IRs apart from manipulating the ADSR of original IR file but it was taken over by Reverberate by introducing dual IR engines plus more control and extra functions.

Altiverb is famous because it is used by pros specially film producers because of hundreds of common and uncommon spaces it offers. Will you use all of them or most of them, first question? Many of us use only two or three tried and tested halls, some spring reverbs, some room and ambiance reverbs. For example my go to is Reverberate and mostly one fusion IR - Live hall. So, depending on how much you use in your 90% of the time in production just ask yourself first that spending 600-700 euro will do you big favour or not? I also sometimes use vastly lauded, extremely inexpensive VNXT EMT 140 impulses for having more dynamics because these IRs are processed by Nebula the great . And when it comes to nebula, none of the fashion icons comes close because, well we know why .

 

Do you guys thinks it is worth it to get Altiverb even if I have Algos (Valhalla Vintage Verb), Analog (Tsar-1), and Adaptiverb for weird things. Is it really that much of a must have? When I really need convo I just use Fruity Convolver, hue hue.

The only 3rd party IR Convolution verb I have and tried is IQ Reverb, but I never get satisfied with it

 

I actually do program.

We don't listen, nor process audio at 2 bit (except specifically for distortion, etc FX). The only thing that matters is what you ACTUALLY process/listen to audio in. 32bit DSP is enough. And with that, there is no problem. Unless you want to do silly things like reverb something at -9001dB or +9001dB.

 

Last edited: Nov 20, 2016

No way. You would know about what I talked, if you did. I'm involved in creating a zero-latency-convolution reverb, btw.

That doesn't make sense. First of all, I said that I'd make it simpler for you to understand what the issue is. But you still didn't get it.

32-bit DSP is enough? Erm, a 32-bit DAW is enough, yes. But to output the best (read it again: the best) 32-bit quality, the plugin actually needs to calculate in 64-bit (that's called internal engine). That has nothing to do with the DAWs audio quality, nor the system or the OS. It is just a totally normal neccessity. I try it one last time with an even simpler image and hope that you can follow the programmer's everyday use of bits.

You have a calculator that is limited to the maximum number of 32. Any higher number will be NIL. Now you type 4 * 8. What will be the result? You see, the two numbers you type are within the set limit of 31, but the result is higher. To prevent this you need a calculator that goes much higher (in this example 31 * 31 = 961 would be ideal). 31 can be presented with 5 bits. If we double the bit count to 10, we can present numbers up to 1023. Oh my god, now it would be able to show all numbers that you can get through multiplication of 5-bit numbers. Isn't that amazing?
There's only one issue. I can only count to 31, and you have to tell me a series of results that go higher than this number. Simple solution: You remap all the numbers to match the range 0-31, so that I can understand the numbers you tell me. And as a result the numbers that have been calculated between 0-961, are now in the range 0-31 (961 becomes 31, 600 becomes 20, etc.). That's a form of interpolation. But the curve they describe is the same as before (just in a lower 0-31 resolution). And that is the important point. Without double the bitdepth there wouldn't be a curve, just a long line at the maximum value, which doesn't represent the original data.

I hope you don't misunderstand this post. If I would just want to behave like a douche I wouldn't write so detailed. The simplicity of the example is really only there to help you understanding the issue. It is not meant to affront your intellect.

 

Just wanted to say I totally like Melda's Multiband Convolution...very good quality sound, also uses quite a bit of computer power, especially when im mixing in 96khz now. but again great stuff i use it all the time on every single track i make these days, thanks.

 

The problem is people are not mixing properly with Vocals an Reverb.
The way I mix a vocal into a reverb is very different than the masses.
Any reverb will sound great but you have to send to it properly.

1. use the Phase Button to have a control over how deep the vocal folds into the Reverb,

Here is how I do it in Reaper

Track 1 Vocal ( COMP, EQ ) make a SEND to Track 2 and make it full send as PRE FADER/POST FX
MASTER PARENT is ON ( So this Vocal track also hits the master Bus)

Track 2 ( Or Bus) Add Reverb and make sure the Reverb is 50% WET/ 50% Dry this is IMPORTANT!!

Go back to track 1 and hit the phase button on and off ..listen. Now move the fader slowly down and up and listen!!

Your Vocal now sits inside the Reverb and not on top of it depending on the slight fader movement
You can now further control your reverb to tweak it and make it unique.

What is happening with the phase button is you are cancelling the dry frequencies from the Reverb along with the Space frequencies that even dry vocals have so what ends up happening is they cancel out. Leaving the Vocal sit inside the space .


If you do a post PAN post FX send the phasing will not work the send must be PRE FADER/ POST FX.


I hate the dry vocals that sound like they are on top of the reverb and not inside it. This is the majority of mixing these days with the silly POST PAN/POST FX sends which I only use for NON VOCALS like Guitars and Drums and others.

 

Last edited: Nov 20, 2016

I think Altiverb is worth having. I don't think it sounds better than Valhalla Room, 2c Aether or The Lexicon Halls and Chambers if you are looking for an orchestral sounding "realistic" verb plugin. I have had Altiverb since the beginning and it has served me well in certain situations. It is a very useful tool but not the best sounding reverb. It is like the junk food of reverbs, it tastes great at first but if it is all you eat, your mixes will get sick. There is a very limited amount of it that sounds good in the finished product. It just builds up and sounds flat, fake and glassy very quickly to me compared to Algo verbs. It tends to sound good in isolation but not good in a big mix with a synth orchestra or even a basic pop band situation. The big spaces are the strong suite, the "captures" of the effects units and toys and weird stuff sound awful and are virtually useless. Altiverb 7 does introduce some really powerful methods to quickly integrate your own impulses from any wav and do some really fun processing in a non-faux reverb way. Especially rhythmic effects.

I agree with the poster that said the factory issued impulses are better quality than the other guys doing impulse reverbs. The factory stuff has a pleasing round deep sound that 95% of other IR plugs I have listened to don't have. All the others seem to accentuate the downside of IRs with the thin aliased brittle frozen character even when they might have some interesting architectural variations. Not there yet I guess.