[tool] WAV melody to PSG conversion

[roudoudou]

inspired from an MSX tool, here is a new tool to convert melody/speech/instrument to PSG frequencies
the tool use a WAV as input, you may select the number of channel, replay frequency, ...
it is possible to create a WAV preview of what you will get on a real CPC
the default output is a listing of register in CSV format, you may choose DMA-list to create an optimised almost ready to use DMA-list
joined to this thread a DSK for Amstrad Plus with some Street Fighter 2 sounds (17K of DMA for ONE MINUTE of speech, cries, ...)
i hope @Targhan will allow some integration with Arkos Tracker for SFX in games with music 
reference: https://msx.org/forum/development/cross-development/convert-wavmp3-psg-melody?page=0
sources in github: https://github.com/EdouardBERGE/wav2ay

[gurneyh]

Great tool, great toy !

Some tests on CPC old
https://rasmlive.amstrad.info/edit/zNAikaHwbFyp9yAe4
https://rasmlive.amstrad.info/edit/aqXEnH33qYPqmrKmw

click on the emulator window to play the sound.

[Targhan]

Very interesting! This could replace some samples in music, or create a new kind of SFX, with the same amount of CPU (but slighly more memory).
I could integrate this to AT2, but not short-term (unless some people urge me to).

[GUNHED]

Thanks a lot! Saves me the work of doing it, had a similar idea. 

[roudoudou]

Very interesting! This could replace some samples in music, or create a new kind of SFX, with the same amount of CPU (but slighly more memory).
I could integrate this to AT2, but not short-term (unless some people urge me to).

[roudoudou]

New release v1.2
audio glitches bugfixes
experimental support of noise register
optimised for size CPC replay
AKI instrument export  
note: this tool is intended to be used by AY power user

https://github.com/EdouardBERGE/wav2ay/releases/tag/v1.2

[GUNHED]

The holy grail of CPC music! 

[SpDizzy]

Thanks so much Edouard for this wonderful tool.

Been toying around for a while with parameters on various input files getting great results on output .wav files.

However, cause I'm using the great Arkos Traker 2 to play music and effects, I wonder:

- Is by chance any possibility to have an output as .aks to use directly on AT2?

As .aki format is limited to 64 lists, even a small speech of 0.25 secs failed to open when tweeking with some parameters for better fidelity with source ('tresh', 'high /low'), chrashing AT2 in the attempt.

I don't want to renounce using AT2 for my productions/tests, so:

- Could be 'dmalist' output data compatible in any way with AT2 .akx output to be used with any AT2 player?

Using 'cpclist' as output I'm in need of a new player as in 'replay_cpc.asm' example, but I even can't get it working because some includes/dependencies are missing like 'ultrasyd_C713.bin'.

Thanks so much both @roudoudou and @Targhan for your time and dedication.

[roudoudou]

this is an example of code, the ultrasyd.bin is a compiled music i used for my tests, just replace the bin with your compiled music or a RET instruction if you do not had one
the player is a proof that the music can be stopped, then sample played, then music continue
if you are a bit lost, i can help you, (maybe mail or discord?)
for AKS it's a non optimal solution for samples, my player use its own packed format
cheers

[SpDizzy]

Wow, that was really a quick response!

Looking in deept at the code now, I can understand how it works (besides I'm not good at all understanding PSG stuff).
So that packed format for the player is just the '-cpclist' output from wav2ay conversion?

PD: I thought 'sample.cpc' was indeed the music. I was terribly clueless. Sorry for the mistake about that. God bless Ultrasyd and all his legacy. A talent of other world.

[roudoudou]

Wow, that was really a quick response!

Looking in deept at the code now, I can understand how it works (besides I'm not good at all understanding PSG stuff).
So that packed format for the player is just the '-cpclist' output from wav2ay conversion?

PD: I thought 'sample.cpc' was indeed the music. I was terribly clueless. Sorry for the mistake about that. God bless Ultrasyd and all his legacy. A talent of other world.

correct

some samples needs to be fine tuned where there is noise involved

as an example, this is the "eight" of Street Fighter 2, i wanted to add some noise in the final "T"

i add +2 to the first value (descriptor) and i add a value at the end of the line (typically 0 for opening noise all channels or #38 to shutdown noise)

once you know that, you can also remove noise when it's wrongly automatically added by the algo!

Code Select Expand

defb 27 ; nombre d'iterations
defb #FC,#04,#00,#9D,#07,#00,#7B,#04,#00,#66 ;  157-04  123-07  102-04
defb #88,#09,#8C,#69,#07,#24 ;  140-09  105-07    36-07
defb #A8,#0C,#7D,#08,#77,#09,#23 ;  125-12  119-08    35-09
defb #A8,#0A,#7E,#0C,#70,#0A,#23 ;  126-10  112-12    35-10
defb #E8,#08,#01,#25,#0D,#91,#0B,#24 ;  293-08  145-13    36-11
defb #E8,#0E,#00,#84,#0A,#22,#0A,#21 ;  132-14    34-10    33-10
defb #88,#09,#F4,#83,#0F,#7D ;  244-09  131-10  125-15
defb #A8,#0B,#7F,#0D,#70,#0A,#20 ;  127-11  112-13    32-10
defb #A0,#09,#7D,#0C,#6D,#1F ;  125-09  109-12    31-10
defb #A8,#0A,#DB,#0B,#6C,#09,#1E ;  219-10  108-11    30-09
defb #20,#DF,#0C,#6E,#1B ;  223-10  110-12    27-09
defb #20,#E2,#0A,#72,#1C ;  226-10  114-10    28-09
defb #A8,#0B,#F1,#08,#76,#0B,#1D ;  241-11  118-08    29-11
defb #A8,#0C,#FB,#0A,#7C,#07,#1F ;  251-12  124-10    31-07
defb #E8,#0B,#01,#08,#09,#80,#08,#1C ;  264-11  128-09    28-08
defb #88,#0A,#0F,#89,#07,#83 ;  271-10  137-09  131-07
defb #A8,#09,#1C,#08,#90,#05,#1C ;  284-09  144-08    28-05
defb #A0,#08,#3E,#07,#96,#1B ;  318-08  150-07    27-05
defb #A8,#06,#61,#03,#90,#00,#1B ;  353-06  144-03    27-00
defb #A0,#03,#B3,#00,#90,#1B ;  435-03  144-00    27-00
defb #80,#02,#78,#90,#1B ;  376-02  144-00    27-00
defb #80,#00,#78,#90,#1B ;  376-00  144-00    27-00
defb #00
defb #00
defb #00
defb #A8+2,#0A,#13,#0A,#88,#0A,#69,0 ;  275-03  136-02  105-02  Add noise
defb #A8+2,#09,#13,#09,#88,#09,#69,#38 ;  275-00  136-00  105-00 Remove noise



[SpDizzy]

some samples needs to be fine tuned where there is noise involved

Thanks so much @roudoudou for detailled info, will take it into account for fine tune.

What I still missing is how to create that compiled music (music with initialisation code).

If we take as example the sample "Fight" of Street Fighter, all we have is raw optimised list of values for the music, obtained as output with '-cpclist' option.

So first call to #C000 for initialise music (where music data is absolutely positioned) will get just 'random' values in there for a joyfull crash, not opcodes.

Code Select Expand

org #4000
run #4000
di

xor a
ld hl,#C713
call #C000 ; init de la musique
So, is it possible to know how that initialisation code looks like? Can't access 'ultrasyd_C713.bin' to take a look.

Thanks so much.

[roudoudou]

What I still missing is how to create that compiled music (music with initialisation code).

oh, i thought it was obvious, sorry!

the compiled music stands for songs of any tracker (Arkos, Starkos, SoundTracker), this is not related to my tool. Only to demonstrate that my sample replay can stop any replay

the ultrasid bin is "default" song provided with Arkos Tracker 2 in directory songs/STarKos of AT2 (i used Fractals because it's louuuuuud)

if you export song as AKG or AKY, both players use "hooks" => that means there is 3 jump at the begining, the first for INIT, the second for PLAY and the third to STOP, this is compatible with old trackers and also when you want to try AKG or AKY with AT2, you keep the same CALL to #C000 / #C003 when compiled in #C000 disregarding the  player your are using. Is it clear?

[SpDizzy]

Damn!! It was really easy 
Perfectly clear, many many thanks!

[Targhan]

How many of you would be interested in an integration with Arkos Tracker? You'd drag'n'drop a WAV and it would automagically use Roudoudou's code to create a special instrument (not limited to any length) to use. I guess this could be used as a normal instrument in the patterns, but also as a sfx.

I'm still dubious to the quality of the samples (many tests I did were not very good), but I'd be happy to have my mind changed .

[Prodatron]

I didn't look at this before, but I would be very interested in it!

[trocoloco]

How many of you would be interested in an integration with Arkos Tracker? You'd drag'n'drop a WAV and it would automagically use Roudoudou's code to create a special instrument (not limited to any length) to use. I guess this could be used as a normal instrument in the patterns, but also as a sfx.

I'm still dubious to the quality of the samples (many tests I did were not very good), but I'd be happy to have my mind changed .

That would be a nice feature, for synth sounds, lazer or even drums & perc would be great as instrument to experiment with, let alone for "samples"  in games

[SpDizzy]

Really interested here of course!!  It would be great to be used as an instrument without that lenght restriction, and marvellous to be exported as .akx

[roudoudou]

 @Targhan let me know if you need something for integration or anything else

[Targhan]

Thanks. I still need to be convinced about the sound quality. If you have wav of before/after samples, I would like to hear that.
Also, I notice in the command line that you can have 3-channel output. Is there a real benefit to this?

[roudoudou]

Thanks. I still need to be convinced about the sound quality. If you have wav of before/after samples, I would like to hear that.
Also, I notice in the command line that you can have 3-channel output. Is there a real benefit to this?

depend on the target. For inspiration, one channel is enough then you can build a better instrument, for voices, 3 channels give strong results regarding the size

here is an example of replay in situation : music playing, then you can press 1,2,3,4,5,6,7,8,9,0 (not from keypad) and listen some SF2 sounds

[Targhan]

Well, I'm sure this could be tweaked and improved, but to be honest, I think it really sounds bad . I wouldn't want that in a SF2 game, it makes absolutely no sense. It sounds like the little dinosaurs in Puzzle Bobble spewing gibberish .

I can still be convinced of course. Because even if it seems technically simple to play such "samples", it would still require a specific implementation in AT (a specific instrument view, export tweaks, specific code for SFX players, etc.). Quite a lot of work. I wouldn't do that if the result isn't worth it.

[roudoudou]

New release v1.3 with WAV Float-32 PCM support

https://github.com/EdouardBERGE/wav2ay/releases/tag/v1.3

And a few fixes which were already in previous daily build

[SpDizzy]

Thanks so much @roudoudou for this update. Really enjoying the tool, very helpfull for someone completely null with music like me