@Jona Sphinx, not sure it is still availabe or not:
http://probing.wikidot.com/sphinx3-to-pure-data
Last blog post from 2017, looks alive: https://cmusphinx.github.io/

I coded a gui to another speech recognition toolkit (mainly for japanase language) (with OSC):
http://www.workinprogress.ca/software/kiku/about/

Cheers~

@Jona Very interesting! I was and still interested in speech recognition in pd. 10 years ago, I did this:

I was trying to sound like a native english speaker, very funny to hear myself talking this way, and not being able to say three properly.

Cheers~

@nicnut Here's how I use it

Exactly, you need to split your main patch in multiple patches: main (trying to be around 80% of a single core), FX (for example) also around 80% (on second core), etc... you get the idea.

There's a trick btw to send midi to pd~ I don't recall exactly what was it, but poke me if you need that too.

Cheers~

@domien

@domien Excellent stuff! Can you describe what pd is doing exactly in this jam?

Cheers~

Good question! But if it's like their last piece (Egregore) it's all pd dsp with gem. Looks like shaders are involve ?

Egregore:

Big fan of chdh.

I fixed a seaboard for a friend and I'm lucky he gave it to me. And now I want to use it in pd... Anyone did the bridge? Will take a look at the patch (thanks @formuls for sharing it).

@jancsika it works! very very cool.