Voice synthesis on ISR

Page 2/27
1 | | 3 | 4 | 5 | 6 | 7

By alexito

Hero (555)

alexito's picture

17-05-2016, 23:55

I have tried using BlueMSX (NMS8245) and the voices sound better using a SCC cartridge for sure.

What is the limit (in KB) for samples?

Great job!!!

By ARTRAG

Enlighted (6275)

ARTRAG's picture

17-05-2016, 23:59

I've just fixed a bug in the PSG version and padded the rom to 16KB because openmsx does not like 5KB roms Big smile
PSG audio should be improved a bit. Try it again if possible

PS
the size is 15 bytes per frames under SCC and 9 bytes per frame on PSG

By alexito

Hero (555)

alexito's picture

18-05-2016, 00:07

Wow! Good Improvement now PSG is almost similar quality to SCC.

Good Job ARTRAG!!!

By [WYZ]

Champion (417)

[WYZ]'s picture

18-05-2016, 00:16

Here we have a sample of ISR Speech: Sky Jaguar

For sure it can be improved but the first step is the hardest.
Many thanks to you ARTRAG for your great coding skills and your enthousiasme.

By JohnHassink

Ambassador (5417)

JohnHassink's picture

18-05-2016, 00:23

[WYZ wrote:

]Here we have a sample of ISR Speech: Sky Jaguar

Hehehe, already nice! :D

By [WYZ]

Champion (417)

[WYZ]'s picture

18-05-2016, 00:46

@JohnHassink Yeah! Smile

And a new demo with a fixed bug: Sky Jaguar 2

By ARTRAG

Enlighted (6275)

ARTRAG's picture

18-05-2016, 07:35

Awesome! Is the voice on the psg ?
Am I wrong or it is not interrupting the psg music ?

Ps Huey is going to develop an encoder for pc so anyone will be free to generate its samples without my matlab scripts

By [WYZ]

Champion (417)

[WYZ]'s picture

18-05-2016, 07:53

ARTRAG, 2 PGS working here.
With a PC tool it will be easiest to find the optimal band pass for the waves.

By ARTRAG

Enlighted (6275)

ARTRAG's picture

18-05-2016, 08:19

Not bad considering that the data rate is about 500 bytes per second

By ARTRAG

Enlighted (6275)

ARTRAG's picture

18-05-2016, 21:11

I've updated the rom adding in the data a double separate encoding for volumes, one linear for SCC, one logarithmic for PSG.
In the initial release I was converting SCC volumes to PSG volumes "on fly", but this was reducing the accuracy for lower levels in the PSG replayer.
Anyway, the PSG sound improves only slightly, because the levels of the encoded tones are almost always in the high range of the amplitude scale. I'm willing to test a 8 tone vocoder using both PSG and SCC at the same time. It should give a data rate of 24 bytes per frame (1,2-1,4Kb/s) and start to became comparable to telephone quality.
WIP... ;)

Page 2/27
1 | | 3 | 4 | 5 | 6 | 7