Reserved for later use.

I have a Maui, I'll do some testing when I have time (next week maybe)

I can contribute to testing when my build with a Tropez Plus is up and running.

I have a Rio packed away somewhere. It supports RAM too (SIPP and I don't have any SIPPs at the moment.) It would also take one day to upload samples because it's a waveblaster type daughterboard.

FGB wrote:

Unlike the Tropez Plus, the Turtle Beach Tropez and the MAUI don't have an effects processor. To enhance the sound, I have the idea to modify existing soundbanks by extracting the .WAV samples, add a reverb through a software (Audacity for example has such function) and reassemble the soundbanks. I don't have any experience with creating / editing soundbanks, so I don't know if it is important that the sample lengths must remain the same or not. In my understanding a good reverb makes the samples a few milliseconds longer. Anyways, a pre recorded reverb would further increase the dynamic of the music experience in DOS games.

I'm not sure adding reverb would improve the overall sound of any bank. As a homegrown sound engineer I know reverb as a tool that works if used in correct place and in correct type & amount. Youtube is full of 'remastered' MIDI old games' soundtracks which are simply rendered using some large soundfont and a huge amount of reverb. The result is usually pathetic. When producing a soundtrack, you usually have to apply some type of reverb to one group of instruments, another type of reverb to another group of instruments, and leave yet another group of instruments dry (that is, without any reverb). If you just pick one type of reverb and apply it to the whole track, usually it would screw things up.

I am not trying to say that Fabian's reverb idea won't work at all - it might. But I believe it will be tricky.

I don't know for sure, but common logic makes me believe that sample lengths should not be fixed in soundbanks. A reverb applied normally would enhance sample length (sometimes not even by milliseconds, but by whole seconds - depends on reverb type). Anyway, that's not the critical problem since even if it happens so that sample length should remain unchanged, samples can be then trimmed at the same length, simply cutting the remains added by reverberation.

FuzzyLogic wrote:

I have a Rio packed away somewhere. It supports RAM too (SIPP and I don't have any SIPPs at the moment.) It would also take one day to upload samples because it's a waveblaster type daughterboard.

There are SIMM to SIPP converters and they work with Rio (I tested that myself when I had a TB Monterey).

bristlehog wrote:

FuzzyLogic wrote:

I have a Rio packed away somewhere. It supports RAM too (SIPP and I don't have any SIPPs at the moment.) It would also take one day to upload samples because it's a waveblaster type daughterboard.

There are SIMM to SIPP converters and they work with Rio (I tested that myself when I had a TB Monterey).

For a motherboard that used SIPP, I converted SIMM modules using SIMM motherboard connectors (I think I got them from Mouser). You install the SIMM in the connector, and then the pins of the connector fit perfectly in the SIPP slots. They were cheap too, I think less that $1 per connector.

FGB wrote:

User "kemot001" claims he did a conversion of a 4MB UltraSound PnP soundbank, converted from a .SF2 soundbank, but the file isn't available for download.

I think I managed to track that one down, here it is:

https://dl.dropboxusercontent.com/u/107843342 … d%20patches.wfb

I can't give it a try now because of lack of time but maybe someone else can.

j^aws wrote:

I can contribute to testing when my build with a Tropez Plus is up and running.

Thank you, that would be cool!

FuzzyLogic wrote:

I have a Rio packed away somewhere. It supports RAM too (SIPP and I don't have any SIPPs at the moment.) It would also take one day to upload samples because it's a waveblaster type daughterboard.

As the RIO uses a different method for uploading samples (via the WaveBlaster header) it is not possible and or practical to upload WFB soundbanks. As you said - IF possible it still would take a day to load the bank 😀
Therefore I excluded the RIO from the list of the potential sound cards.

bristlehog wrote:

I'm not sure adding reverb would improve the overall sound of any bank. As a homegrown sound engineer I know reverb as a tool that works if used in correct place and in correct type & amount. Youtube is full of 'remastered' MIDI old games' soundtracks which are simply rendered using some large soundfont and a huge amount of reverb. The result is usually pathetic. When producing a soundtrack, you usually have to apply some type of reverb to one group of instruments, another type of reverb to another group of instruments, and leave yet another group of instruments dry (that is, without any reverb). If you just pick one type of reverb and apply it to the whole track, usually it would screw things up.

I am not trying to say that Fabian's reverb idea won't work at all - it might. But I believe it will be tricky.

I don't know for sure, but common logic makes me believe that sample lengths should not be fixed in soundbanks. A reverb applied normally would enhance sample length (sometimes not even by milliseconds, but by whole seconds - depends on reverb type). Anyway, that's not the critical problem since even if it happens so that sample length should remain unchanged, samples can be then trimmed at the same length, simply cutting the remains added by reverberation.

Well I think many wavetable cards only have global effects with a effects processor that is used on the output of the wavetable. So the result is still far away from the possibilites you mentioned like applying effects on certain instruments or ensembles. I think the Ensoniq SoundScape is a good example for pre-recorded reverb.
It's just an idea to do the same to make the music a bit more dynamic when used on the Maui or Tropez. My main concern is the question if one can just batch run all samples and add the exact same amount of reverb or if one has to adjust everything sample per sample. Also I have no clue how to apply the reverb on instruments that use multiple voices.
Do you own a Turle Beach ICS card? I would highly appreciate your advice and/or participation in this project.

stamasd wrote:

I think I managed to track that one down, here it is:

https://dl.dropboxusercontent.com/u/107843342 … d%20patches.wfb

I can't give it a try now because of lack of time but maybe someone else can.

Hi,
thanks for hunting it down 😀 How did you find the soundbank?

I did try the soundbank on my MAUI. It starts loading the bank but reports an error at SAMPLE #452 (of 511 total samples, IIRC). If I try to playback a midi file it sound terrible off with missing instruments.
Maybe the file is broken? It's 6.300.000 bytes big, don't know if that is supposed to be the correct size.

FGB wrote:

I did try the soundbank on my MAUI. It starts loading the bank but reports an error at SAMPLE #452 (of 511 total samples, IIRC). If I try to playback a midi file it sound terrible off with missing instruments.
Maybe the file is broken? It's 6.300.000 bytes big, don't know if that is supposed to be the correct size.

What I understand from the thread: Turtle Beach Soundfonts ??? / WEMP player where the soundfont was first mentioned is that it was converted from the original Gravis Ultrasound Patch set (5.6 MB in size) to SF2 and then converted to WFB. Going directly from a SF2 to WFB would perhaps lead to better results. And besides, both Gravis and Turtle Beach used versions of the Eye & I productions Voice Crystal soundbank, so the difference in sound wouldn't be stellar. A better idea is to buy the V4 of the Voice Crystal soundbank, already in SF2 format, and convert that one.

Hopefully when my Maui arrives I can get it to fit my system (got a super tiny retro system build around an Gravis Ultrasound Ace), so that I can play and contribute to this project.

FGB wrote:

I did try the soundbank on my MAUI. It starts loading the bank but reports an error at SAMPLE #452 (of 511 total samples, IIRC). If I try to playback a midi file it sound terrible off with missing instruments.
Maybe the file is broken? It's 6.300.000 bytes big, don't know if that is supposed to be the correct size.

Try downloading it again, you may have a corrupt file. The file size is 6,481,276 bytes.

(I didn't find the file anywhere, I made it myself. I found the .sf2 bank that was the origin of it on a ftp site following hints from your original post, then converted it with Awave studio - which I purchased many years ago when I was playing with soundfonts for the SB Live; the price wasn't so obscene back then. The trail of my investigation led me first to https://web.archive.org/web/20120309114542/ht … oldscardemu.htm and from there to ftp://fm:fm@atomu.ath.cx:6872/Gravis%20Ultrasound.sf2 )

@Rawit:
I would also expect the soundbank to sound somewhat similar to the default soundbank in the MAUIs ROM but it sounds totally different, with almost all instruments totally off. Maybe it is a conversion error or a problem with instruments that use multiple voices. Your suggestion to directly start with a "proven to be working" .SF2 soundbank is good as it is always the best way to reduce potential sources of errors and / or problems.

@stamasd:
Thanks for the details, I checked the size, but it is the same. So the file is complete but the soundbank seems broken when I load it in DOS.
Yeah, I would love to try AWAVE myself but I find it a bit much money just for trying a SF2 to WFB conversion.
Can you maybe be so nice and try to convert one of the well known 4MB or 8MB .SF2 soundbanks to the WFB format? I would highly appreciate your effort.

I have actually tried to convert other soundfonts but haven't found yet one that converts cleanly apart from the one above. All other .sf2 files I have tried have too many samples for the WFB format and would need to have samples removed to "fit" the WFB container. I can't find anywhere a description of the WFB format to find out how many samples I would need to cut. That goes even for much smaller soundfonts - I've tried 1MB and 2MB files, and get into the same problem.

I've started making manually a list of all the samples in the GUS soundfont and will try to compare with lists of samples from other soundfonts, but it will take time because each contains hundreds of samples, and awave doesn't have any sort of automation for this type of activity. I literally have to write everything down on paper. I don' think I can effectively do it for more than one other soundfont, so please suggest which one I should concentrate on.

That's an interesting observation! I think the sample quantity is limited to 512 samples, at least I think I remember this number from the DOS Initialization programm when it reports 354 of 512 samples in use or something like this. Maybe this is the limit of the WFB bank.

OK, I will suggest a SF2 soundbank. Thanks for your effort!

Does it struggle with the total amount of samples or with the amount of samples per instrument? Because lots of soundfonts that I've encountered, including small ones, are General Midi + extra, like GS or XG. It could be very well that WFB only supports GM level 1/2 instruments and nothing more. Converting the 128 GM instruments only + the drumkit might do the trick regarding the amount of samples.

The Roland and Yamaha soundfonts from Deemster, posted earlier on this forum, are both < 512 waveforms according to Awave 11 Demo version. Could these be candidates for an conversion?

The GUS soundfont has 321 waveforms (samples) total. Another soundfont that fails to convert has 319 samples. 🙁

(edit) oh wait, for this one it actually complains about the number of instruments, not the number of patches. Let me investigate.
(edit2) Hm, I don't get it. The file that fails to convert has the exact same number of instruments and drum kits as the one that can be converted. And fewer samples overall. Yet awave complains of too many instruments.

Rawit wrote:

The Roland and Yamaha soundfonts from Deemster, posted earlier on this forum, are both < 512 waveforms according to Awave 11 Demo version. Could these be candidates for an conversion?

Not sure about the Yamaha soundfont. The Roland SC-55 Preset one fails.

So I've tried editing the "Roland SC-55 Presets.sf2" soundfont to remove all of the non-GM instruments. It now has only GM, but 495 sampls. Awave doesn't want to convert it to WFB because of too many samples (not too many instruments).

I've run out of time to fiddle with this today. I'm attaching the edited soundfont, perhaps someone with the trial version of awave can go through it and tell me which samples I should remove when I have time to work on this again.

Also, if someone has a known good and working WFB file, please upload it.