silly little python pipeline for live translation. (whisper-stream didn't work for korean)
microphone input -> whisper.cpp -> llama.cpp -> translated output
- running whisper.cpp server (example:
$WHISPER/build/bin/whisper-server -m $WHISPER/models/ggml-large-v3-turbo.bin --host 127.0.0.1 --port 8081 -t 8 --language ko -pp -pr -sow -mc 768) - running llama.cpp server (examople:
llama-server -m gemma-3-12b-it-GGUF/gemma-3-12b-it-Q4_K_M.gguf -ngl 48)