Voice

[TheSonicRetard]

To the OP's original question: I'd much rather have text I can quickly read ans scroll through in an RPG than each line being voiced. A game like JB Harold Murder Club on the TurboGrafx does have each and every line of text voiced and there is a lot of text in the game. If you leave the voices on the game really drags. Turn them off and you're free to play at your own speed. The Intellivision's voice module was synthesized speech, like Speak and Spell or crappy text-to-speech programs. Cool for the time, kind of laughable today. Though I always thought that the Sports Talk games on the Genesis sounded like Speak and Spell, but according to something I read somewhere the voices are all digitized.

You sure the voices on the Intellivision weren't digitized? Different games had different voices.

[SEGA.GENESIS1989]

The one thing that I noticed about Technosoft's voice samples was how muffled the sound was. You had to really listen to the words to understand what they were trying to convey.

One of these days I need to hook up my SEGA Genesis to speaker system. Neighbor to the right and neighbor to the left, prepared to plug your index fingers in each of your ears!

[kool kitty89]

This has always bugged me, the fact that 16-bit video games like Genesis had no voice audio, but instead made you read through scrolling pages of text, especially on RPG's and certain platformers. Can it be explained why Intellivision had its voice module in the early 80's with computerized voices, but Genesis and SNES could not put audible voices into games some ten years later?

Cartridge space. Lots of voice samples would require huge ROMs to store the samples. See Super Street Fighter II.

He made the express point to mention the Intellivoice though . . . which used speech synthesis rather than samples and thus used tiny amounts of ROM by comparison. (which is also why most speech in arcade games into the mid 80s was synthesized and not sampled)

And as to the reason for not using synth speech in games . . . for 1 there's the issue of hardware (needing an add-on or having speech synth built-in), handling synthesis in software (CPU intensive and requiring suitable software encoder/decoder to be designed, and then the inherent limited quality of synthesized speech.

However, the YM2612 natively supports speech synthesis decoding in hardware (with moderate software assistance that should be easily handled by the Z80), but very few if any games made use of that for whatever reason.

Synthesized speech isn't limited to that primitive mechanical sounding speech either, but could potentially improve with more complex/higher bitrate streams as well (it is effectively a form of lossy compression, so quality can be improved to some extent) though at some point 8-bit digital samples (or ADPCM) will be definitively superior.

I'm not sure, but at a guess, it's possible that Sega provided very little or no support for this feature of the YM2612.

You sure the voices on the Intellivision weren't digitized? Different games had different voices.

They're synthesized, again, like contemporary arcade speech. (such synthesis could partially re-create individual voices or accents to a fair extent -Atari Star Wars is one of the better examples with synth speech sounding somewhat like their actual voices from the film)

It's definitely a different process than just using fixed samples of different words or vowel/consonant sounds and stitching them together. (like the Sports Talk games)

Edit: specifically, for the Atari Star Wars speech, a TMS5220 chip was used. See: http://en.wikipedia.org/wiki/Texas_I...C_Speech_Chips

[Christuserloeser]

However, the YM2612 natively supports speech synthesis decoding in hardware (with moderate software assistance that should be easily handled by the Z80), but very few if any games made use of that for whatever reason.

Didn't Tiido experiment with speech synthesis a few years ago? I think I remember something like that.

[GeckoYamori]

In retrospect I prefer the dayts when text dialog was the way to go in RPGs. There are two major problems: first one is that voice acting simply requires that much more work and it's often not possible to have dialog on the same scope as older games when everything is expected to be voiced. This usually means you end up with watered down writing. Just compare the dialog in Oblivion to Morrowind and this becomes blatantly evident. Second is that if the acting quality is not good enough, it usually hurts immersion more than it adds to it (Chaos Wars or whatever it's called for PS2 is an extreme and hilarious example).

Doesn't help how companies like Bioware insist on even the player character being voiced these days.

[Bones Justice]

I agree that voice synthesis would be too CPU-intensive. What I remember about speech synthesis during that era even for computers is that it tended to "take over" whenever it was playing. The quality of speech wasn't very good, either, compared to sampled speech. There are several Genesis games that offer a lot of sampled speech, like the SportsTalk games or Jammit!, but they probably aren't enough for a role-playing game.

[tomaitheous]

If it's for an RPG like the OP stated, what does it matter that it might hog some CPU resource? Also, the intellivision and other systems had the synth voice module and a much slower processor than the Genesis. So it's not like it wouldn't have eaten as much cpu resource relatively speaking. Also noted was the minimal amount of space required for the synth voice compared to sampled recording output.

OP, that's a great question and definitely a missed opportunity IMO. Now, sure people wouldn't care. But back in the day, this would have been incredible to have on a cart for an RPG. And on the Genesis, it was doable fairly easy since there are input lines on the cart port.

[havok666]

Phelios has some dialogue. A Couple short sentences are voiced, although they sound like you're underwater.

[kool kitty89]

If it's for an RPG like the OP stated, what does it matter that it might hog some CPU resource? Also, the intellivision and other systems had the synth voice module and a much slower processor than the Genesis. So it's not like it wouldn't have eaten as much cpu resource relatively speaking. Also noted was the minimal amount of space required for the synth voice compared to sampled recording output.

OP, that's a great question and definitely a missed opportunity IMO. Now, sure people wouldn't care. But back in the day, this would have been incredible to have on a cart for an RPG. And on the Genesis, it was doable fairly easy since there are input lines on the cart port.

The YM2612 itself should be able to assist with some types of speech synthesis, though for later-gen stuff, it probably wouldn't cut it quality-wise compared to on-cart coprocessing possibilities. (without added hardware, speech synth could still make more sense than PCM/ADPCM/CVSD within the practical bitrates/space limitations of the time, but with an embedded on-cart DSP, MPEG-like compressed audio might be possible)


OTOH, ~22 kbps compressed audio (ie 11 kHz 2-bit or 22 kHz 1-bit ADPCM/CVSD) might have worked reasonably well too (possibly better than using the YM2612's sythesis capabilities) while still offering low enough bitrates to allow several minutes of speech to fit in a fair chunk of late gen carts. (165 kB per minute, so just over 3 minutes in 512k) ~6 kHz 4-bit PCM (or ADPCM,or DPCM) would use similar bitrates and require less optimized Z80 coding for playback, but usually shouldn't sound nearly as good for speech.
For early-gen games, any decent amount of speech (more than a few seconds) would only be practical using the sythesis capabilities (or awkward sample-based speech synthesis routines, building words/phrases out of a limited set of speech samples -as in the Sports Talk games).

Both of those latter options (CVSD/ADPCM and Yamaha-accelerated synth) would be possible without any add-on hardware, just the Z80+YM2612. (outputting to the 8-bit DAC in the case of software decoded ADPCM/CVSD)

[Tasuke]

Cartridge space. Lots of voice samples would require huge ROMs to store the samples. See Super Street Fighter II.

also see SNK NEO-GEO MVS/AES...

[Tasuke]

The one thing that I noticed about Technosoft's voice samples was how muffled the sound was. You had to really listen to the words to understand what they were trying to convey.

One of these days I need to hook up my SEGA Genesis to speaker system. Neighbor to the right and neighbor to the left, prepared to plug your index fingers in each of your ears!

my Genny is patched in through a truly high-resolution Hi-Fi system.
everything passing through it sounds rich and full bodied in the extreme.

regardless, any and all of those awesome TECNO-SOFT Genny samples tend to remain quite troublesome indeed to properly comprehend.
it does indeed help, in no small measure, if you realize, and try adapt to, the fact that T-S was using a female japanese voice actor, attempting to call out weapon names/etc. in English,
through a rather thick and difficult Japanese accent. watching and carefully listening to a lot of Subtitled Anime helps develop and hone this skill to no end.

then, of course, there is the legendary (IMO) case of the "GOOD LUCK" sample at the start of THUNDER FORCE II.
even on a highly resolving Hi-Fi such as my own,
i can never manage to understand the sample as anything besides VOICE 1; "MUMBLE/EMPHYSEMA/MUMBLE-MUMBLE" VOICE 2; "KNOTTER, NIG-KNOCK"

[Christuserloeser]

Turns out the YM2612 is capable of speech synthesis:

http://www.sega-16.com/forum/showthr...l=1#post502717

[Stef]

Yeah, and it was already mentioned earlier in this post :

However, the YM2612 natively supports speech synthesis decoding in hardware (with moderate software assistance that should be easily handled by the Z80), but very few if any games made use of that for whatever reason.

The CSM mode for channel 3 permits the formant speech synthesis, unfortunately a few emulator handle it correctly as we have few documentations on this feature but this is something i want to play with at some point.
I think it's still quite limited in quality compared to digitalized speech but could permit free space speech