Automatically convert Zoom meetings to text

Zoom is a wonderfully practical tool for online meetings. And since the meeting is held on the computer, it can also be easily recorded. There is a “Record” button in the toolbar. The recording is automatically saved on the computer on which “Record” was pressed. This is also possible with free accounts.

If you have a paid account, you should turn off “Cloud recording” in the account settings beforehand, then the recording will also be saved on the computer and automatically converted to an mp4 file after the meeting has ended.

This file can be automatically converted into text using the f4x speech recognition program.

Attention! Some pitfalls for speech recognition in online meetings

With speech recognition, the better the recording, the better the result.

However, this is not trivial in practice. The built-in microphone on the laptop works quite well for short meetings. However, it is extremely unsuitable for speech recognition. The person speaking is relatively far away from the microphone, you can usually hear reverberation. This has a negative effect on detection accuracy.

What can you expect?

Here we have recorded a short conversation via Zoom and had it translated into text using f4x speech recognition. Person 1 used a good headset, person 2 used the microphone built into the laptop. Both people are well recognized, but person 1 with headset is much better, interjections from person 2 are sometimes not recognized at all.

Example

#00:00:01-2# All right, Mr. Müller, I’ll start the recording. I hope you agree with this. <unk> yes, must be included now, Blech agrees. Yes, we’re taking that on board now too. This is simply safer for us to document if you could explain the topic of today’s meeting to us again <unk>.

#00:00:24-5# That

#00:00:26-4# Expansion of the A 49. O

#00:00:30-2# Yes, all right, that’s an exciting topic. <unk>. I actually wanted a more neutral topic.

#00:00:38-5# There may be different opinions, yes, of course

#00:00:49-5# <unk>

#00:00:55-4# <unk>. What do you see them thinking about, <unk>, lunch, <unk>. Has that changed in the course of the Krona era?

#00:01:07-7# Yes, wonderful. Thank you very much. Lunch in its wonderfully informal theme. We can work with thank you very much <unk>. I think that’s enough. And that was it. Yes, thank you very much, <unk> gladly and goodbye to many at lunchtime. Yes,

Conclusion: Zoom transcript as a protocol

As you can see: This is not a precise transcript, but the text is very suitable for recognizing topics.

Thanks to the time markers contained in the transcript, you can simply jump to the relevant point and, if necessary, listen to the original recording again – e.g. with the transcription software f4transkript.

So if you are planning to create an automated transcript of the meeting, you should ensure that a) all participants use good microphones and b) the conversation contains as little simultaneous speech as possible.