The Ur-Quan Masters Home Page Welcome, Guest. Please login or register.
Did you miss your activation email?
November 14, 2019, 02:50:25 am
Home Help Search Login Register
News: Paul & Fred have reached a settlement with Stardock!

+  The Ur-Quan Masters Discussion Forum
|-+  The Ur-Quan Masters Re-Release
| |-+  Technical Issues (Moderator: Death 999)
| | |-+  Speech Sound Quality Clean Up Mod
« previous next »
Pages: [1] Print
Author Topic: Speech Sound Quality Clean Up Mod  (Read 3348 times)
RealityHacker
Zebranky food
*
Offline Offline

Posts: 8



View Profile
Speech Sound Quality Clean Up Mod
« on: January 03, 2009, 08:47:39 pm »

Hiya!

I've really liked playing UQM so far - it's awesome! One thing that have bothered me though is the sound quality of the voice acting, of which most (but not all) is very good. After doing some testing with the files to improve the quality (reducing 11050Hz noise and other static, reducing overtones, restoring clips and removing clips etc.), I've come to the conclusion that it's viable to moderately (sometimes greatly) enhance the sound. This without compromising the original sounds or boosting current file sizes. The idea, of course, would be to eliminate elements of the sound that interfere with the actual content to bring forth the *juicy* parts. However, in dealing with over 2100 individual sound files, this is potentially a rather daunting project (but hurray to the wonders of sound processing automation!).

Now I have some questions:

1. Is there any similar project in the works (not counting music remixes and such)? I have not found any searching the forums.

2. Is there uncompressed versions of the speech or is the OGG files compressed in a completely lossless manner? Wouldn't be clever to start out with files that already have been reduced in quality, now would it?

3. Would you who happen to read this be interested in listening to processed sound and give feedback so that I can concentrate on processing?

That's all!
Logged
meep-eep
Forum Admin
Enlightened
*****
Offline Offline

Posts: 2847



View Profile
Re: Speech Sound Quality Clean Up Mod
« Reply #1 on: January 03, 2009, 09:05:39 pm »

1. There is no project to enhance the quality of the existing samples, as far as I'm aware of.
2. The voice files on the 3DO CD had their own compression (not sure if it was lossless). This compression wasn't very good, and we could save a lot of space by recompressing these files as .ogg. The samples themselves were of low quality, and encoding them as .ogg didn't significantly make them worse. Still, if you're going to reprocess them, it would be better to work on the .wav files you get from decoding the .abx files from the CD, and re-encode them afterwards. The Subversion repository contains tools to decode the .wav files.
3. I would be interested in hearing a few of them as an example of what you managed to do with them.
Logged

“When Juffo-Wup is complete
when at last there is no Void, no Non
when the Creators return
then we can finally rest.”
RealityHacker
Zebranky food
*
Offline Offline

Posts: 8



View Profile
Re: Speech Sound Quality Clean Up Mod
« Reply #2 on: January 03, 2009, 09:30:45 pm »

1. Cool.

2. It's not likely the 3DO files were lossless as these methods wasn't that common back then and the space and performance requirements were kinda harsh. The repacked files (lossy) are OGG Vorbis, 16 Kbps bitrate (which is very low and might introduce some issues), mono with a sampling rate of 11050 Hz. As for the extracted files, I'm not sure how to compile the tool and then extract the data (I also don't have the original 3DO CD). If someone could help me out with this, it'd be great! And rewarding!

3. Of course.
Logged
RealityHacker
Zebranky food
*
Offline Offline

Posts: 8



View Profile
Re: Speech Sound Quality Clean Up Mod
« Reply #3 on: January 04, 2009, 05:30:23 am »

Here are some examples, featuring Captain Fwiffo:

http://drop.io/uqmvoiceremaster

Note that the "after" versions of these files are based on the compressed OGG files shipped in the UQM voice pack. This means that the "after" versions are rerepacked versions of the original sound files (which isn't a good thing). The "after" versions uses the aoTuVb OGG codec which gives better results than the standard at low bitrates, which is something I doubt that the usual OGGs do.

Also note that while the removal of noise, clips, clicks together with frequency balancing, loudness maximizing etcetera makes the sound clearer that also enhances other issues, such as microphone plosives (puffs) as well as the occasional bad acting.

When listening, please turn off any equalizers in your media player and amplifier.

Feedback is appreciated!  Grin I might try my luck on another voice as well (just to test - it's a lot of work) but in order to get this done proper I'd need the .wav extracts from the original 3DO CD. Unfortunately, I have not been able to figure out how to do that. Any volunteers? I know you want to!
Logged
Novus
Enlightened
*****
Offline Offline

Gender: Male
Posts: 1938


Fot or not?


View Profile
Re: Speech Sound Quality Clean Up Mod
« Reply #4 on: January 05, 2009, 12:44:23 pm »

2. It's not likely the 3DO files were lossless as these methods wasn't that common back then and the space and performance requirements were kinda harsh.
Looking at the decoder, it would seem that the compression used is a variable bit-rate variant of ADPCM, compressing 8-bit linear PCM to frames consisting of 0, 1, 2 or 4 bits/sample and 7-bit resynchronisation points. There is no way this process is lossless. Sampling as 8-bit LPCM at 11025 Hz doesn't do much good to an audio signal either.

In any case, there were two ways to get the 3DO audio data:
  • Extract the 3DO CD file system (see here). Use the conversion tool in sc2/tools/abx to convert to WAV.
  • Download the WAV files from the UQM CVS repository. Being a version management repository, it contained all the releases of UQM, including the initial TFB release before Vorbis compression was introduced. Unfortunately, the WAV files seemed to have been dropped in the move from CVS to SVN.

The resulting files are identical (checked by converting everything to signed raw LPCM) except for:
  • druuge/druug011: The UQM copy contains a failed take which has been removed in later versions of UQM and in 3DO SC2.
  • umgah/*: The 3DO version of the Umgah voice seems to have been through some dynamic range compression that doesn't exist in the UQM version.
To make it easier on you, I've arranged so you can download the UQM original WAVs (as a ZIP containing FLAC files) here. Note that this link is likely to be temporary (university web server; can't leave large personal files there indefinitely).
Logged

RTFM = Read the fine manual.
RTTFAQ = Read the Ur-Quan Masters Technical FAQ.
Megagun
Enlightened
*****
Offline Offline

Gender: Male
Posts: 580


Moo


View Profile
Re: Speech Sound Quality Clean Up Mod
« Reply #5 on: January 12, 2009, 11:42:23 pm »

Wow, these sound a lot nicer and crisper than the "before" ones. Good job. Smiley
Logged
RealityHacker
Zebranky food
*
Offline Offline

Posts: 8



View Profile
Re: Speech Sound Quality Clean Up Mod
« Reply #6 on: January 13, 2009, 06:03:01 pm »

Why, thanks (to both of you).  Wink

I have downloaded the audio files - just was I was looking for! I'll get back to you when progress has been made. I think that the best way to go about this is to do the clean up incrementally, one voice at a time, as they need quite a lot of individual tweeking. By releasing them in the drop (above) and notify in the thread, perhaps people will give the files a listen and give feedback. One could of course also do a Mod pack even though not all voices are included.
Logged
Razorback
Frungy champion
**
Offline Offline

Gender: Male
Posts: 99


We are all *happy campers*


View Profile
Re: Speech Sound Quality Clean Up Mod
« Reply #7 on: April 18, 2009, 07:18:03 pm »

What kind of media playback device handles .OGG files?
Logged

meep-eep
Forum Admin
Enlightened
*****
Offline Offline

Posts: 2847



View Profile
Re: Speech Sound Quality Clean Up Mod
« Reply #8 on: April 18, 2009, 08:23:30 pm »

I think this is what you are looking for: http://wiki.xiph.org/VorbisHardware. Quite a lot these days, apparently.
Logged

“When Juffo-Wup is complete
when at last there is no Void, no Non
when the Creators return
then we can finally rest.”
RealityHacker
Zebranky food
*
Offline Offline

Posts: 8



View Profile
Re: Speech Sound Quality Clean Up Mod
« Reply #9 on: April 18, 2009, 10:28:04 pm »

Hi again.

I've been fiddling a bit with this lately and wanted to know what you think the optimal export format would be. The current one is very low (and also uses an outdated encoder):

Format                           : Vorbis
Bit rate                         : 16.0 Kbps
Channel(s)                       : 1 channel
Sampling rate                    : 11.025 KHz
Writing library                  : libVorbis 1.0 (UTC 2002-07-17)

My suggestion is:

Format                           : Vorbis
Bit rate mode                    : Constant
Bit rate                         : 48.0 Kbps or 32.0 Kbps
Channel(s)                       : 1 channel
Sampling rate                    : 32.0 KHz or at least 22.5 KHz
Writing library                  : aoTuV b5c  (UTC 2008-12-15) or later (it rocks at low-rate encoding)

If this format would cause a major performance hit or cause too large a download, maybe it isn't a good idea to up it this much. However, not going a bit higher in quality than the current level would in practice undo most of the work (it would just sound blurred, possibly static or garbled).

Input, anyone?
Logged
meep-eep
Forum Admin
Enlightened
*****
Offline Offline

Posts: 2847



View Profile
Re: Speech Sound Quality Clean Up Mod
« Reply #10 on: April 19, 2009, 01:42:36 am »

The larger size isn't as much of an issue anymore as it was when the project was started. Besides, we can give people the choice.
Logged

“When Juffo-Wup is complete
when at last there is no Void, no Non
when the Creators return
then we can finally rest.”
AngusThermopyle
*Smell* controller
****
Offline Offline

Gender: Male
Posts: 304


A paranoid android.


View Profile
Re: Speech Sound Quality Clean Up Mod
« Reply #11 on: April 23, 2009, 05:02:01 pm »

Hey, I agree about these sounding a lot crisper. Keep up the good work!
Logged
Pages: [1] Print 
« previous next »
Jump to:  


Login with username, password and session length

Powered by MySQL Powered by PHP Powered by SMF 1.1.21 | SMF © 2015, Simple Machines Valid XHTML 1.0! Valid CSS!