>>91538122Tbf the LLM also has to generate the text (4 words a second), but I guess that's pretty fast. I reckon Zen could speed it up massively if she streamed the text from her STT into the TTS, rather than having the TTS start inference once the text is fully transcripted.
>>91538150>filter AI can't keep upYeah, it can? I'm certain a classifier could inference much faster than the voice.
