Record mode

This is referring to the ChatGPT Record mode available on the MacOS client.

It allows up to two hours of audio to be recorded and at the end, it creates a transcript with summary.

Initially it only provides the summary, although I think you can instruct it to provide the transcript as well by default, but I haven’t seen that working.

However, in the summary, it typically provides references to timestamps, and clicking any of those will cause the full transcript to open on the side.

In theory, after a recording you can ask questions in the chat about it, but I have found that to not work reliably.

What I have had work, at least in the context of wanting to get feedback on the quality of a practice run-through of a presentation, is to copy the transcript into a file and upload that into a separate conversation. In my case, since I was the only speaker, I also did a find and replace to remove my name from all of the timestamp markers. I also added angle brackets around the timestamps to make them more separate.

It could probably be rigged up to hear the contents of a voice or video call as well, but you’d probably need something like Loopback to set it up. It also makes it very clear you should let people know they are being recorded.