Yeah, a secret is that the synthesis stuff (the subtractive synth and friends) work in the old audio engine! I don’t highlight that since it works better in the audio mixer, and I want something fun to talk about for audio mixer. They’re based on an object which is used in both the new and old audio engine (USoundWaveProcedural). This object, incidentally, was written ages ago, primarily to support in-engine VOIP and wasn’t really thought out. But it was a hidden gem of awesomeness that I turned into a full-blown synthesis engine.
The audio mixer does synthesis much better than the old – every synth callback is actually in an async task and is treated like any other source decode. This should allow you to be able to run more synthesizers without risking underruns in the audio mixer.
Also, with the audio mixer you get source effects and submix effects, which are really cool.