Big Bird's Hide and Speak sample compression
Moderator: Moderators
That's how the Digitalker family of codecs works, by treating voice as periodic over the short term. But in addition to periodism and 2-bit delta coding, it employs two other non-obvious (at the time) techniques. It changes the phase of each period sample to all cosines so that it can use a ping-pong loop (2x saving), and it silences the half of that sample that has the least energy (another 2x saving). That's how games using MX squeeze voice into 4 kbps or smaller at the expense of a bit of "robot voice". The developers of BBH&S probably didn't want the "robot voice" and were willing to spend a lot of bits to eliminate it.rainwarrior wrote:The ghostbusters thing doesn't sound like fixed bit compression to me. If I had to guess what's going on there, I'd say it cuts the sound into short segments and replaces each segment with a single cycling waveform that is a close match.
Yes, and the PCM sample had to be stored logarithmically because of this.tokumaru wrote:But it's played back using the square wave channels, right?Dwedit wrote:Sonic the Hedgehog for Game Gear uses 4-bit PCM audio for the SEGA sound.
Now I am getting a bit of a better understanding of how these PCM-based speech engines work. I've got the jist of how the sample playback works, but it is animating the character's mouth that I am wondering about.
Of course, accurate lip synching is not an overall necessity, considering that the sample playback requires 100% CPU attention. I mean, in Sesame Street Countdown, when the Count says, "'Ah, 'ah, 'ah! Find [number]," his mouth only opens three times. I'd love to find out how Hide & Speak manages to keep Big Bird's mouth moving during smaple playback.
Of course, accurate lip synching is not an overall necessity, considering that the sample playback requires 100% CPU attention. I mean, in Sesame Street Countdown, when the Count says, "'Ah, 'ah, 'ah! Find [number]," his mouth only opens three times. I'd love to find out how Hide & Speak manages to keep Big Bird's mouth moving during smaple playback.
Moving the beak during sample playback is as easy as updating a couple entries in a display list, waiting for vblank, and then blasting the display list to the PPU through $4014. This probably happened at the breaks between concatenated waves, seeing as all the voiceless consonants are stored as separate waves.
The rate at which the graphics have to be animated is pretty low. I'm sure there's some time between the output of consecutive samples to update some sprites and/or background tiles. In a mapper like the MMC3, that can bankswitch small portions of CHR-ROM, it would be even easier.NEStegg wrote:I'd love to find out how Hide & Speak manages to keep Big Bird's mouth moving during smaple playback.
- rainwarrior
- Posts: 8732
- Joined: Sun Jan 22, 2012 12:03 pm
- Location: Canada
- Contact:
Most of the samples are broken up in such a way that facilitate animation. Here's an alternative dump with a pop between every individual sample.
http://rainwarrior.ca/projects/nes/Sesa ... %20div.wav
Notice, for instance, that "hide and" are together as a single sample, and as you might expect, he does not close his mouth between them in-game. Not sure why the alphabet letters are so broken up; I don't think there's any animation during them, really.
Also, this game is MMC1, if anyone's wondering. No special timing hardware was involved, it's pretty much all just cycle counting.
Dwedit, are you sure about Ultimate Stuntman? The NSF version is constantly playing DPCM samples in every track.
http://rainwarrior.ca/projects/nes/Sesa ... %20div.wav
Notice, for instance, that "hide and" are together as a single sample, and as you might expect, he does not close his mouth between them in-game. Not sure why the alphabet letters are so broken up; I don't think there's any animation during them, really.
Also, this game is MMC1, if anyone's wondering. No special timing hardware was involved, it's pretty much all just cycle counting.
Dwedit, are you sure about Ultimate Stuntman? The NSF version is constantly playing DPCM samples in every track.
Last edited by rainwarrior on Mon Sep 07, 2015 8:34 am, edited 1 time in total.
Yes, Ultimate Stuntman definitely uses raw PCM during gameplay as well.rainwarrior wrote:Dwedit, are you sure about Ultimate Stuntman? The NSF version is constantly playing DPCM samples in every track.
Download STREEMERZ for NES from fauxgame.com! — Some other stuff I've done: fo.aspekt.fi
- Doommaster1994
- Posts: 162
- Joined: Mon Jun 21, 2010 10:23 pm
- Location: Seattle, WA
- Contact:
I know Michael Riedel over the net if you guys want me to ask him anything. Though he never replies to my messages anymore, but I can try. I remember mentioning this game to him and he didn't have a lot to say about it. I also know one of the programmers if you're interested, though again, he didn't have a lot to say about the game either.
- rainwarrior
- Posts: 8732
- Joined: Sun Jan 22, 2012 12:03 pm
- Location: Canada
- Contact:
Re: Big Bird's Hide and Speak sample compression
http://www.glitterberri.com/developer-i ... f-sunsoft/
Apparently Mozer ESS was used in Mito Koumon 1 and 2 for the Famicom as well.
Apparently Mozer ESS was used in Mito Koumon 1 and 2 for the Famicom as well.