What if the audio effect is:
replyTake an audio stream, apply STT, interpret the result as an algorithm, run the algorithm, take its output, apply TTS, then send that audio stream to the output.
The several seconds of latency would be fun.
replySounds like it would be susceptible to how they’d reverse engineer LLMs using distillation.
reply