Voice synthesis on ISR

Page 26/28
19 | 20 | 21 | 22 | 23 | 24 | 25 | | 27 | 28

By NYYRIKKI

Enlighted (5396)

NYYRIKKI's picture

22-11-2019, 22:32

Yes! Now at least for me this gives much better HW/result value. Good job!

By [WYZ]

Champion (417)

[WYZ]'s picture

22-11-2019, 23:27

Thank you NYYRIKKI, I really appreciate your words.

By ARTRAG

Enlighted (6276)

ARTRAG's picture

23-11-2019, 00:23

Speech is as good as before, greetings! Your player with voice effects should find place in games and demos.
Great work!

By [WYZ]

Champion (417)

[WYZ]'s picture

23-11-2019, 00:52

Only 3 SCC channels. Smile

This is part of your work too.
And complex SFX are also a new sound universe to discover.

By jltursan

Prophet (2190)

jltursan's picture

23-11-2019, 19:14

All in all seems like black magic to me oO

Wouldn't it be great to add this to the AGD engine?

By ARTRAG

Enlighted (6276)

ARTRAG's picture

23-11-2019, 20:43

Great idea! The problem is to have a light stand alone encoder for voice.
Anyone willing to port to C the voice encoder?

By Grauw

Ascended (8508)

Grauw's picture

27-11-2019, 00:25

For a singing voice, since it has a single pitch, if you take the IDFT at the fundamental frequency to produce the waveform, do you need more than a single SCC channel? Since an SCC waveform can convey up to 16 harmonics (ignoring stepping noise) it seems to me like it should be able to reproduce the formants with relative accuracy… What are the additional channels used for?

So you would scan over an array of fundamental frequency + waveform for each frame, 2040 bytes/s. The waveforms can perhaps be shared when their DFT is similar to reduce the storage requirements.

I wonder how it would sound, seems like it should be fairly good due to the high rate and accurate reproduction of the pitch, but maybe it’ll be a bit autotune-ey :). Still cool. Good for Cher :P.

By [WYZ]

Champion (417)

[WYZ]'s picture

27-11-2019, 11:20

Something like Dvik SCC demo - Leila K? (but this method is nothing related with ISR samples...)

https://www.youtube.com/watch?v=SvCHnrNKV8Q

By ARTRAG

Enlighted (6276)

ARTRAG's picture

27-11-2019, 12:26

No, the idea is to use the wave form to better match the spectral maxima without using other channels
At the time I have fiddled around this concept without any acceptable result.
One of the problems was that usually the speech had more formants at not multiple frequencies.
The advantage would be that you leave the other channels free.
I will return on the subject if have time.

By ARTRAG

Enlighted (6276)

ARTRAG's picture

01-12-2019, 14:53

I've fixed a bug in my old code and using one pitch seems to sound not so bad
wip on this ...

Page 26/28
19 | 20 | 21 | 22 | 23 | 24 | 25 | | 27 | 28