It's typical to use a special STT model (audio in only), which will not be able to clean up afterwards. If you are using an LLM for the STT part, you're leaving

It's typical to use a special STT model (audio in only), which will not be able to clean up afterwards. If you are using an LLM for the STT part, you're leaving stuff on the table