Source and SoundData

[MachineCode]

I have a question about the audio system. What I want to do is make a SoundData buffer of a fixed length and then synthesize samples into that using some algorithms. This needs to be wrapped into a Source so it can be played or looped.

If I modify the SoundData with SoundData:setSample, will this change what gets played the next time? There does not seem to be an update for Source to reflect changes in the SoundData buffer.

Do I need to create a new Source from the SoundData to effect the changes? If so, how much overhead is that likely to create? The wiki has a warning about repeatedly calling newSource, but does that apply when using in memory SoundData buffers (as opposed to reading files or activating decoders)?

[pgimeno]

I suggest you take a look at this library which is already doing that: https://github.com/nucular/sfxrlua

[zorg]

Unfortunately, even if you modify the SoundData object, it won't have any effect on an already playing source unless you load it into the Source again.
But, i did experiment with realtime sound generation myself, and currently, lpghatguy's QueuableSource module would be something you should try in my opinion

I also linked my own thing so you can see how i got it to work, it's p. ugly though, and probably unoptimized, but at least it mostly works!

[MachineCode]

OK thanks for that. I will take a look at your project - it looks like you went down a similar track, but it does not look simple.

The problems I see are first, that after modifying the SoundData, you need to create a new Sound object and discard the old one - which is probably quite expensive if done often.

The second one is that there is no built in way to queue a second Sound to play after the first. That means that trying to set up ping-pong sound buffers in a loop is not going to be viable. Playing or looping a segment of a Sound doesn't seem to be available, although it should be possible.

I need to do some more reading.

[zorg]

You're welcome.

Technically, it's enough to modify the SoundData, and queue that up with the QSource, if using that; but it could be done faster of course, using pure FFI arrays.

Also, OpenAL does have a built-in way to queue data constantly, though löve itself doesn't (expose it nor implements it in any way), at the moment.

[bobbyjones]

You're welcome.

Technically, it's enough to modify the SoundData, and queue that up with the QSource, if using that; but it could be done faster of course, using pure FFI arrays.

Also, OpenAL does have a built-in way to queue data constantly, though löve itself doesn't (expose it nor implements it in any way), at the moment.

Yeah that's the reason we can't have Issue #1 solved already.