[wordpad25]

How did you get the latency so low?

Is it going through your own backend or directly connects to OpenAI for transcription?