usholanb / asr_chat

records voice, refines transcripts with LLM based on transcript history, translates. Live with redis bullmq

Translates voice live.

Stack: Whisper model/API, LLM via https://vellum.ai to refine transcripts and translate and keep history for more context translation quality

requirements

pip install -r requirements.txt

torch

pip install torch torchvision torchaudio

install whisper

pip install git+https://github.com/openai/whisper.git

install whisperx

pip install git+https://github.com/m-bain/whisperx.git

run in this order

python async_processor.py python async_recorder.py