3 ms·

Does the audio go direct to chatgpt, or through a transcription model first? I could see it being a bit inaccurate if the transcription silently corrects my er