(Insert stupid Blog name here)

I'd like to see an Mac AtariVox app, more than a GUI version of KrokCom. The current KrokCom works just fine (although a GUI version would be nice, someday).

 

One nice feature to add, that PhraseALator (likely) doesn't do, is generate 2600-ready code. I would think this wouldn't be too hard to do.

 

The key though, is in being able to access all of the sounds easily, and be able to quickly manipulate them (and their parameters) to make the voices sound somewhat-natural. That trial-and-error is what takes forever.

The key though, is in being able to access all of the sounds easily, and be able to quickly manipulate them (and their parameters) to make the voices sound somewhat-natural. That trial-and-error is what takes forever.

That's precisely why I thought that adapting John's 16-char demo to Bob's AtariVOX source would be useful, as mentioned here. The only issue is that the demo doesn't display quite right on real hardware, but I think John is working on that.

 

EDIT: answered by own question - the PhraseAlator can interface to the Avox via USB. So I assume the USB interface is a custom device. Will it ever be sold, and could it (in theory) allow management of the Savekey EEPROM data?

According to the manual, PhraseALator interfaces with the AtariVox, so you're using the real hardware for playback.

The key though, is in being able to access all of the sounds easily, and be able to quickly manipulate them (and their parameters) to make the voices sound somewhat-natural. That trial-and-error is what takes forever.

That's precisely why I thought that adapting John's 16-char demo to Bob's AtariVOX source would be useful, as mentioned here. The only issue is that the demo doesn't display quite right on real hardware, but I think John is working on that.

 

EDIT: answered by own question - the PhraseAlator can interface to the Avox via USB. So I assume the USB interface is a custom device. Will it ever be sold, and could it (in theory) allow management of the Savekey EEPROM data?

I believe that Richard H. is working on creating the USB device and David Galloway (djmips) is writing the software for it.

 

There also are, floating around (I have one), a serial interface that goes between the PC and the AVox. And I have used this to dump the EEPROM - in fact, I just used it a month or two ago when debugging Elevators Amiss.

I'd like to see an Mac AtariVox app, more than a GUI version of KrokCom. The current KrokCom works just fine (although a GUI version would be nice, someday).

agreed

One nice feature to add, that PhraseALator (likely) doesn't do, is generate 2600-ready code. I would think this wouldn't be too hard to do.

I like this idea

The key though, is in being able to access all of the sounds easily, and be able to quickly manipulate them (and their parameters) to make the voices sound somewhat-natural. That trial-and-error is what takes forever.

cd-w gave me a little 4K program so you can compose speech using a Cuttle Cart or Supercharger rather than relying on the USB interface. I also made an AtariVox header file for it that makes it easier to string allophones together. Let me know if this is useful.

 

Blog Post with Attachment

 

What I do is speak it myself the way I'd like it to sound, then adjust the parameters word by word until it matches how I say it. It works well for me. Riding the pitch is very important in naturalistic delivery. Although the AtariVox does a poor job with certain allophones (p in particular is almost inaudible) when you humanize the phrasing, it is more intelligible because your mind fills in the gaps.

 

The turnaround time between tweaking the bytes and hearing the change with the Cuttle Cart using this method is like 10 seconds. It's not that bad. It's like anything else, though. The more time you spend on it, the better the result.

 

BTW, I think text to speech is for chumps. If you want it to sound good you have to hand pick every allophone. There are just too many ways to pronounce things. TTS will always be a little off. I mean, with direct control over the allophones you can approximate accents like Corcodile Dundee or Shrek. It's just so much more powerful.

I guess now would be the time to post my hack of Supercat's 16-char demo that sends the on-screen data to the Avox (in hex.) Set the numbers with the joystick, and press Reset to talk. Note that this does not work quite right on real hardware (number display is off) but in Stella w/Avox it should work (though I have not tested it this way.)

Riding the pitch is very important in naturalistic delivery. Although the AtariVox does a poor job with certain allophones (p in particular is almost inaudible) when you humanize the phrasing, it is more intelligible because your mind fills in the gaps.

 

See here for a brilliant example of how delivery is more important than phonetic precision. With the right wording, the lesson of the story makes sense:

 

Mural: Yonder nor sorghum stenches shut ladle gulls stopper torque wet strainers.