@hanzjwermhat
Best way to do this is probably to deploy a Whisper + DeepSeek (or other open source) model on a server than call through a web socket api connection. Phones have no probable running whisper for transcription (although it’s actually a bitch to get the model running) but back and forth inference even the smallest models the watch can’t run on device.
0 reply
0 recast
0 reaction