#Google "Coral" A.I accelerator for TTS/STT and/or llm for home assistant voice?

1 messages · Page 1 of 1 (latest)

glossy adder
#

Is there a integration to use coral for processing the LLM in the ram with this device? Or will it only work with nvidia and cuda?

Or can it also use intel UHD 630 for the processing my any chance?

frail kettle
#

TPU won't work for LLM.

sly mist
#

presently for whisper stt you need a nvidia cuda GPU, even a cheap used one like a GTX 1070 or 1080 will do the job.

#

tts like piper can run on a solid CPU. There's GPU accelerated versions of it (I run that), but honestly with a good CPU the time difference is pretty small.

muted mantle
#

can piper run on another machine to offload that from my rpi?