Can P2 speak?
knivd
Posts: 102
This is a question which interests me a lot! With a good bunch of smart I/O and eight cores, can the P2 be made to generate natural sounding speech? For example, the Amazon Echo uses one of the consumer Mediatek CPUs (ARM A-something) and Linux (Ugh... 
), and with that it manages to deliver excellent quality TTS.
Is it theoretically possible that something similar could be achieved in bare metal with a P2?
Comments
Amazon Echo does basically nothing on-device except recognize its wakeword.
There's a voice synth object for P1, but its difficult to actually get recognizable speech out of it. On P2 it could be higher quality, but the biggest issue is getting good parameter envelopes from plain text.
It'd be neat to try out Amazon Polly... Looks like free for one year...
There is a speech generator example for P1. Maybe it can be translated to a P2.
There was something called narrator.device on the Amiga about 35 years ago. It ran on a plain 7MHz 68000 with a 8 bit DAC and DMA (see Paula discussion). So this should definitively be possible on the P2.
Reading plain english text it wasn't performing very well. But that's because the character to phonetic translation is generally quite hard for this language. The main problem is to get the emphasis and pitch modulation right so that it sounds natural and not like a robot.
Certainly something much simpler would / should work. Pre recorded (your own voice) key words (Voice of America uses 130 English words) saved as a WAV or similar file on the P2 SD card would be ideal for me. A bit choppy maybe and inflection off, but a start.
A list of spoken letters of the alphabet, numbers 1 to 9, eleven and on, twenty, thirty and on, hundred, thousand, million and on - common if, then, the, is /are - there are numerous lists out there.
To read back. define a command "Speak" followed by your sentence or command sequence.
As in:
PUB Speak
Lookup, "P2", "can", "speak", "A", "B", "C"
Just a thought.