Syntazer - A speech synthesizer for Famicom

This is an archive of a topic from NESdev BBS, taken in mid-October 2019 before a server upgrade.
View original topic
Syntazer - A speech synthesizer for Famicom
by on (#238257)
Hello!
I'm here to show you one of my projects I'm currently working on.

Syntazer - A speech synthesizer for Famicom with Polish accent

Like any other speech synthesizer, it converts an input text into a speech (no duh).
The algorhythm is very similar to these used in other simple speech synthesizers, like Black Box v8.1 for Commodore 64. It converts an input text into phones and then plays in series audio samples that correspond with these phones. Because of a noticable small number of the character input, a support for pages have been implemented so you can fit more text. It's still WIP, so maybe it's not that impressive as it seems to be. I'm also going to add a support for the Family BASIC keyboard. Also, apologies for making it sound too beep-boop-robotic.

When it's fully done, I'm going to save the PRG code into an EPROM and put it on a PCB board with a circuit similar to these in UNROM/UOROM boards and test it on a real hardware.
Re: Syntazer - A speech synthesizer for Famicom
by on (#238261)
It sounds good. That must be using $4011 output, right? It is robotic, but very clean sounding. Dougeff and myself have both made DPCM-based speech synths based on a recorded voice, makes it sound somewhat more human, but voice samples rarely sound great with DPCM encoding.

Besides the audio, text to phonemes is pretty impressive. I'd looked into doing it in English and it seems like one would almost need to use a dictionary lookup, as just following basic rules would quickly lead to a lot of wrong pronunciations. Was that aspect of it any easier in Polish?
Re: Syntazer - A speech synthesizer for Famicom
by on (#238262)
Memblers wrote:
That must be using $4011 output, right?
Correct! It's steraming raw 7-bit audio data through a $4011 register.

Memblers wrote:
Was that aspect of it any easier in Polish?
In Polish, a single letter can be treated as a single phoneme, it's not like in English or other languages where a single letter have multiple phonemes, e.g. "I" is basically pronounced as "AI" which are two different tones combined together.
That means doing a conversion from text to phonemes in Polish makes it a ton easier.
There are some exceptions, though. Some single phonemes can be made from two letters, e.g. "SZ" is pronounced as "ʃ".