We used @GroqInc to build https://t.co/e8kiR8fxCZ and do real time text-to-speech -> llm -> speech to lip sync. And the best part, it was fast enough that we didn't have to even use streaming to shuttle requests.
@paulg Agreed! When we started working on https://t.co/yUGpOJAA2v(a podcasting social platform with built-in editing tool) we were using our own AI model.The emergence of Whisper changed the game for us. Now it generates transcripts that are 99% accurate & doesn't require human editing.