Voice synthesis on ISR

Página 9/30
2 | 3 | 4 | 5 | 6 | 7 | 8 | | 10 | 11 | 12 | 13 | 14

Por ARTRAG

Enlighted (6406)

imagem de ARTRAG

31-05-2016, 08:48

This is a windows version of the encoder
https://sites.google.com/site/voicesynthesisontheisr/home/te...

- unrar the files in a directory
- place your .wav files in the subdirectory wav\
- open a command window and go in the directory where sccLOFI3.exe is
- run from commandline sccLOFI3.exe
- enjoy the new rom
or
- report which DLLs are missing from the package ;-)

Por ARTRAG

Enlighted (6406)

imagem de ARTRAG

31-05-2016, 09:04

@Hit
I should have moved the noise to a weaker psg channel, does it improve?

Por Huey

Prophet (2651)

imagem de Huey

31-05-2016, 09:56

ARTRAG wrote:

This is a windows version of the encoder
https://sites.google.com/site/voicesynthesisontheisr/home/te...

Thanks!

Quote:

- report which DLLs are missing from the package ;-)

I'm missing mclmcrt8_0.dll

Por ARTRAG

Enlighted (6406)

imagem de ARTRAG

31-05-2016, 10:26

Por ARTRAG

Enlighted (6406)

imagem de ARTRAG

31-05-2016, 12:39

Download the Windows 32-bit version of the MCR for R2012b from the MathWorks Web site by navigating to

http://www.mathworks.com/products/compiler/mcr/index.html

For more information about the MCR and the MCR Installer, see
Distribution to End Users in the MATLAB Compiler documentation
in the MathWorks Documentation Center.

Por Metalion

Paragon (1155)

imagem de Metalion

31-05-2016, 16:05

ARTRAG wrote:

Techniques for Harmonic Sinusoidal Coding

Nice read indeed !
Wink

Por ARTRAG

Enlighted (6406)

imagem de ARTRAG

01-06-2016, 10:17

I've submitted to the downloads DB the current version of the stand alone encoder for windows.

The instructions are those above, let me recall them:

Unpack the content of the .rar in a subdirectory.

In order to run it, you have to install the Windows 32-bit version of the MCR for R2012b from the link above (MATLAB Compier Runtime environment).
The MCR package is huge, so I would suggest first to try to download only mclmcrrt8_0.dll from here
https://sites.google.com/site/voicesynthesisontheisr/home/te...
and to place it in the path of the encoder.
@Huey I changed the version of the DLL at the link, this time it should be the right one.
If this does not work, install the MCR files.

The encoder is the file sccLOFI3.exe and will process all the wav files in the wav\ subdirectory.
The files will be encoded in alphabetic order, thus from basic
?usr(0) will play the first sample, ?usr(1) the second etc etc
The encoder will also try to generate a rom by compiling the new data and the asm player in the package.
The compiler will call sjasm.exe, I added it to the package in case you do not have it already in your path.

If the data do not fit in the 256KB rom the assembler will report error, anyway all asm sources are included so you can tune the number of pages according to your needs.
Thanks to the ascii 8 mapper you can have up to 255 8KB pages (2MB rom), that corresponds to about 24 minutes of audio at SCC quality.
Just change 31 to 255 any time you see the line
page 1..31
and
defpage 1..31,0x6000, 0x2000
in the ASM files and in the TXT files

I think that current version should allow to add voice to any msx project, even to basic games.
;)

Por [WYZ]

Champion (421)

imagem de [WYZ]

03-06-2016, 14:28

Great! Hannibal

Por ARTRAG

Enlighted (6406)

imagem de ARTRAG

03-06-2016, 16:51

I think there is a lot to improve, also on psg only.
Now the frequency resolution used in the extimation is very coarse, about 60Hz.
It should be 3Hz Tongue
Moreover the noise is activated on the first channel in strength.
It shoud instead be activated only if unvoiced sounds are d detected and only on the tone falling in the unvoiced band
Add that the base frequencies should be estimated in 50 - 450hz..
Anyway, having only 3 square waves, I do not know if is better to get all the harmonics but at low amplitudes or less harmonics but at closer amplitude....

Por hit9918

Prophet (2895)

imagem de hit9918

04-06-2016, 08:31

the PSG adds tones of 3x 5x 7x frequencies of the square wave.
if you also look at the 3x frequency in the diagram,
like volume = (volume(1x freq) + volume(3x freq)) / 2
maybe something good happens.
the 3x doesn't need to be found as a peak, just take its value.

or maybe a good formula is
like volume = (volume(1x freq) + volume(3x freq) * 3) / 2
because the 3x is in 1/3 amplitude in a square.

so, if the sample is a square, the volume will get 1, but if it is a sine then the volume will get 0.5.
maybe the PSG gets more dimmed in those places where it makes the extra tones.

Página 9/30
2 | 3 | 4 | 5 | 6 | 7 | 8 | | 10 | 11 | 12 | 13 | 14