Tila - free beta reverb (was: very-beta reverb plugin...)
- KVRAF
- Topic Starter
- 7890 posts since 12 Feb, 2006 from Helsinki, Finland
From page 7:
Version 0.9.2:
- stereo M/S pattern now with cardioid middle (was essentially figure eight)
- low/mid and mid/high crossover frequencies now adjustable
- subtle one-pole lo/hi cut filters for gentle shaping of the extremes
- improved the basic EQ code very slightly for better band interaction
Copying the updated info from page 5 here..
Introducing version 0.9.1:
- Redesigned most internal core parameters for hopefully better response
- Added M/S decomposition to input signals (instead of simple mono-mix)
- Repositioned outputs on M/S field for hopefully better stereo image
- Improved decay equalization for better band separation
- Added matched output equalization to help shape early balance
- Added simple input diffusion to slightly soften the initial echoes
- Roomsize now displays smaller values for a reverb of similar size.
...and...
- Now ships with a non-SSE version for older computers.
----
In other words, this one got some more parameters, much fancier stuff going on under the hood, and it hopefully sounds better (or at least not much worse) than the previous versions.
It does have slightly different character, so it'd be happy to hear comments if somebody wants to compare the original with this new one, and you are advised to keep a copy of the older version if you really want to keep that sound; on the other hand I believe this should be able to replace the old version quite happily in almost all cases.
CPU use unfortunately went slightly up; it should still be quite acceptable, but I'll see if I can bring it back to what it was, but with all the new features/changes that's going to need a bit more work. The SSE version should be faster of the two anywhere it runs. PLEASE report any sudden spikes in CPU use with the non-SSE version. It should not cause any, but I might have missed something somewhere.
Assuming most people don't find it considerably worse sounding than the older versions, I'll probably keep the core engine more or less intact from now on. Might add a couple of more shaping controls for fine tuning at some point, but otherwise this is probably pretty close to feature complete.
---- Original First Post Follows:
Ok, yesterday I decided that since all free reverbs suck (tm) I should really try writing a basic reverb to see how hard it actually is to write a reverb that does not suck very badly (I wrote a reverb for Valo but that one does kinda suck, even if it gets the job done for Valo). So sometime afternoon yesterday I started writing a new plugin.
Unfortunately, after having used most of yesterday writing something that was basically an extended version of Valo's reverb, I realized that the design I had would either require DAYS of tweaking, or I'd better come up with something else.... so I basically deleted the whole processing code, and then last night wrote something else...
And as of this writing, time is 07:39 (as I finally press submit) here, and I'm not sure what to do with the resulting thing. It hasn't crashed so far, so I guess it isn't totally unstable (but what can you say about something that's only some hours old). I'm also getting tired enough, that I'm not sure if my ears can tell what's a good reverb anymore, so I though I'd drop it here (it's supposed to become a free plugin anyway); maybe someone can try it and tell if it sounds like something that could possibly be worth some more work.
- The response isn't totally smooth (there's a bit of ringing), but I'm not feeling like any more trial and error right now.
- The real mid-band time is quite heavily dependant on low/high-band times, which could be improved, but I'm not feeling like doing the math right now.
- There is no attempt whatsoever at doing any sort of proper early reflections (maybe later).
- Time and size are independent, but obviously small size with long time (or the reverse for that matter) won't sound very great. Oh and don't trust the numbers with the size as they are not scientifically derived or anything.
- Obviously I didn't magically come up with a GUI.
- The name is Finnish (surprise) and can mean at least "enclosed space", "room/space for something", "farm" and "state of something" (and I probably forgot something).
- The "pan" control is simply for making my life easier when using the thing as an insert-effect, and does NOT pan the reverb (just the dry signal, before it's fed into the reverb; this way I don't need to use a separate panning plugin or send-track in order to do the panning before the reverb).
Please, if you try this, drop a short note in the thread (or PM if you're shy) saying what you think about how it sounds! And most importantly HOW MUCH DOES IT SUCK??
Download here: http://www.signaldust.com/files/Tila.zip (updated)
(requires a processor that can deal with SSE, I'll do a non-SSE build in a day or two when I have to time to put all the necessary calls to kill denormals in there; edit: does not apply anymore).
Version 0.9.2:
- stereo M/S pattern now with cardioid middle (was essentially figure eight)
- low/mid and mid/high crossover frequencies now adjustable
- subtle one-pole lo/hi cut filters for gentle shaping of the extremes
- improved the basic EQ code very slightly for better band interaction
Copying the updated info from page 5 here..
Introducing version 0.9.1:
- Redesigned most internal core parameters for hopefully better response
- Added M/S decomposition to input signals (instead of simple mono-mix)
- Repositioned outputs on M/S field for hopefully better stereo image
- Improved decay equalization for better band separation
- Added matched output equalization to help shape early balance
- Added simple input diffusion to slightly soften the initial echoes
- Roomsize now displays smaller values for a reverb of similar size.
...and...
- Now ships with a non-SSE version for older computers.
----
In other words, this one got some more parameters, much fancier stuff going on under the hood, and it hopefully sounds better (or at least not much worse) than the previous versions.
It does have slightly different character, so it'd be happy to hear comments if somebody wants to compare the original with this new one, and you are advised to keep a copy of the older version if you really want to keep that sound; on the other hand I believe this should be able to replace the old version quite happily in almost all cases.
CPU use unfortunately went slightly up; it should still be quite acceptable, but I'll see if I can bring it back to what it was, but with all the new features/changes that's going to need a bit more work. The SSE version should be faster of the two anywhere it runs. PLEASE report any sudden spikes in CPU use with the non-SSE version. It should not cause any, but I might have missed something somewhere.
Assuming most people don't find it considerably worse sounding than the older versions, I'll probably keep the core engine more or less intact from now on. Might add a couple of more shaping controls for fine tuning at some point, but otherwise this is probably pretty close to feature complete.
---- Original First Post Follows:
Ok, yesterday I decided that since all free reverbs suck (tm) I should really try writing a basic reverb to see how hard it actually is to write a reverb that does not suck very badly (I wrote a reverb for Valo but that one does kinda suck, even if it gets the job done for Valo). So sometime afternoon yesterday I started writing a new plugin.
Unfortunately, after having used most of yesterday writing something that was basically an extended version of Valo's reverb, I realized that the design I had would either require DAYS of tweaking, or I'd better come up with something else.... so I basically deleted the whole processing code, and then last night wrote something else...
And as of this writing, time is 07:39 (as I finally press submit) here, and I'm not sure what to do with the resulting thing. It hasn't crashed so far, so I guess it isn't totally unstable (but what can you say about something that's only some hours old). I'm also getting tired enough, that I'm not sure if my ears can tell what's a good reverb anymore, so I though I'd drop it here (it's supposed to become a free plugin anyway); maybe someone can try it and tell if it sounds like something that could possibly be worth some more work.
- The response isn't totally smooth (there's a bit of ringing), but I'm not feeling like any more trial and error right now.
- The real mid-band time is quite heavily dependant on low/high-band times, which could be improved, but I'm not feeling like doing the math right now.
- There is no attempt whatsoever at doing any sort of proper early reflections (maybe later).
- Time and size are independent, but obviously small size with long time (or the reverse for that matter) won't sound very great. Oh and don't trust the numbers with the size as they are not scientifically derived or anything.
- Obviously I didn't magically come up with a GUI.
- The name is Finnish (surprise) and can mean at least "enclosed space", "room/space for something", "farm" and "state of something" (and I probably forgot something).
- The "pan" control is simply for making my life easier when using the thing as an insert-effect, and does NOT pan the reverb (just the dry signal, before it's fed into the reverb; this way I don't need to use a separate panning plugin or send-track in order to do the panning before the reverb).
Please, if you try this, drop a short note in the thread (or PM if you're shy) saying what you think about how it sounds! And most importantly HOW MUCH DOES IT SUCK??
Download here: http://www.signaldust.com/files/Tila.zip (updated)
(requires a processor that can deal with SSE, I'll do a non-SSE build in a day or two when I have to time to put all the necessary calls to kill denormals in there; edit: does not apply anymore).
Last edited by mystran on Fri Jul 31, 2009 9:15 am, edited 4 times in total.
-
- KVRAF
- 1702 posts since 26 Feb, 2008
this is amazing on short settings!!
judas priest!
longer reverb times not quite as good. but those small spaces...wow. This is a keeper. Didn't work in FL Studio for me (I am still using 7 though).
I like it the more I use it though! In short - it does not suck. small spaces = very very good (short reverb times I mean)
I will be most definitely be using this on my current project...
judas priest!
longer reverb times not quite as good. but those small spaces...wow. This is a keeper. Didn't work in FL Studio for me (I am still using 7 though).
I like it the more I use it though! In short - it does not suck. small spaces = very very good (short reverb times I mean)
I will be most definitely be using this on my current project...
Snare drums samples: the new and improved "dither algo"
- KVRAF
- Topic Starter
- 7890 posts since 12 Feb, 2006 from Helsinki, Finland
Glad you like it, and even more so if you think the short times are good. Hmmh.rifftrax wrote: longer reverb times not quite as good. but those small spaces...wow. This is a keeper. Didn't work in FL Studio for me (I am still using 7 though).
As for FL7, I actually devel on FL but this time forgot to test on FL7 (still got it installed, using FL8 now though). It could be that I was a bit lazy, and you really need VST2.4 host, 'cos I didn't do the backwards compatibility stuff (kinda forgot).
Will need to add proper exports (which is probably the problem here) and the old-style process calls (which some other older host might care for). I'll see if I can do those later today, actually.
Anyway, thanks for comment, will have to look into improving the longer times.
-
- KVRAF
- 1702 posts since 26 Feb, 2008
and I mean that longer reverb times aren't as good just from a viewpoint of comparing to the simply incredible short times. The longer times still sound fantastic! If you could bring them to the same par as the small room emulations possible with it - there would be a serious gem! I mean, it literally sounds better than very realistic impulse responses in convolution reverbs on short times...
Snare drums samples: the new and improved "dither algo"
- KVRAF
- Topic Starter
- 7890 posts since 12 Feb, 2006 from Helsinki, Finland
Uploaded a new version that now happily works in FL7 (tested) and probably other older hosts.
Have to see what I can do about the longer times (and bigger spaces)...
Have to see what I can do about the longer times (and bigger spaces)...
-
- KVRAF
- 1702 posts since 26 Feb, 2008
i look forward to it. this is an interesting development to be sure...mystran wrote:Uploaded a new version that now happily works in FL7 (tested) and probably other older hosts.
Have to see what I can do about the longer times (and bigger spaces)...
what a pleasant surprise it was to stumble upon this in the first place.
Snare drums samples: the new and improved "dither algo"
-
- KVRAF
- 1669 posts since 4 Nov, 2007
-
- KVRAF
- 1669 posts since 4 Nov, 2007
it does not open in energy xt. seems like it does not have output in the modular view.
-
- KVRian
- 995 posts since 25 Apr, 2005
Hi mystran - this does not suck at all
I have had some excellent results on my acoustic guitar using the short settings. Longer settings sound a little bit weaker but are still easily good enough to use. It is very CPU efficient as well
I have looking into different reverb techniques myself recently and this doesn't sound like the typical Schroeder/Moorer reverberator. Are you using an FDN?
I'm looking forward to seeing how far you can take this
I have had some excellent results on my acoustic guitar using the short settings. Longer settings sound a little bit weaker but are still easily good enough to use. It is very CPU efficient as well
I have looking into different reverb techniques myself recently and this doesn't sound like the typical Schroeder/Moorer reverberator. Are you using an FDN?
I'm looking forward to seeing how far you can take this
-
- KVRAF
- 1702 posts since 26 Feb, 2008
yeah, after comparing this one to ariesverb/glaceverb/reaverbate - it is evident that it has a very very different sound somehow. It sounds 'expensive' - like an eventide algorithm or something. can't put my finger on it.Caco wrote: I have looking into different reverb techniques myself recently and this doesn't sound like the typical Schroeder/Moorer reverberator. Are you using an FDN?
Snare drums samples: the new and improved "dither algo"
- KVRAF
- Topic Starter
- 7890 posts since 12 Feb, 2006 from Helsinki, Finland
Schroeder reverbs are pain to design (no chance I'd come up with a good sounding one overnight) and generally don't sound that great even when sounding good. I personally do not see myself writing that kind of reverberator ever again (I tried a long time ago).Caco wrote: I have looking into different reverb techniques myself recently and this doesn't sound like the typical Schroeder/Moorer reverberator. Are you using an FDN?
Yeah, I'm using an FDN... but then again any lossless network of delaylines can be seen as an FDN... the trick is simply to come up with good matrix (the set of delays is less important actually, and easier to fix anyway).
- KVRAF
- 9453 posts since 17 Sep, 2002 from Gothenburg Sweden
Just did a quick 2 minute test on some well recorded drums. Did'nt crash and did'nt suck. There was some ringing on some settings that could do with some improvement plus some sort of top end attenuattion would be nice.
Not bad,not bad at all.
Not bad,not bad at all.
-
- KVRian
- 995 posts since 25 Apr, 2005
Yep I have found the same thing with Schroeder reverbs, they are quick and easy but they just don't sound that nice. I have only had a quick play around with FDNs so far with reasonable results but I have not come up with anything as nice as this onemystran wrote:Schroeder reverbs are pain to design (no chance I'd come up with a good sounding one overnight) and generally don't sound that great even when sounding good. I personally do not see myself writing that kind of reverberator ever again (I tried a long time ago).Caco wrote: I have looking into different reverb techniques myself recently and this doesn't sound like the typical Schroeder/Moorer reverberator. Are you using an FDN?
Yeah, I'm using an FDN... but then again any lossless network of delaylines can be seen as an FDN... the trick is simply to come up with good matrix (the set of delays is less important actually, and easier to fix anyway).
It is working okay in eXTv1.4 for me, not tried eXT2 thoughRottweiler wrote: it does not open in energy xt. seems like it does not have output in the modular view.
-
- KVRist
- 205 posts since 12 Apr, 2005 from The Netherlands
I haven't tested it extensively but tried it on some acapella's and instrumentals (as i don't have any other samples on this pc right now) but from what I heard: I love it! I'll definately keep an eye on the progress of this plug!
Ps. I tried it in a demo version of reaper 2.43 and it worked fine, I'm gonna try it later on my audio PC in Cubase Studio 4.1.3
Ps. I tried it in a demo version of reaper 2.43 and it worked fine, I'm gonna try it later on my audio PC in Cubase Studio 4.1.3