Otter.ai has released a plugin for Google Meet which allows participants to get a live transcription of the meeting, saving the hassle of taking meeting notes by hand.
The mind-blowing Otter.ai accurately transcribes what people are saying in meetings as they’re speaking. Within a second or two, the words someone has spoken appear on screen, effectively providing live subtitles of what’s being said by all participants.
At the end of the meeting, participants can be sent a full transcript of the meeting, allowing them to search for particular keywords and listen back to key parts of the meeting, without having to wade through the entire recording.
Participants can also add highlights to the meeting transcript as the words appear, allowing them to refer back to key points at a later time. Comments can also be added live, allowing someone to mark out action points from the meeting, for example.
Catching up with Zoom
Otter.ai’s live transcription service has been available for Zoom users for some time. It’s now a premium service, costing $20 per user per month, with a one-month free trial.
For an introductory period, the Google Meet service will be free. Users simply have to install the extension for the Google Chrome browser (which should also work with Microsoft Edge and other Chromium-based browsers) and click on the extension during a Google Meet to begin transcribing. This too will eventually become a premium service, Otter.ai’s CEO and founder, Sam Liang told me.
How is it so accurate?
There are many transcription services out there, but I’ve yet to come across one as accurate as Otter.ai. It’s by no means perfect - it will stumble over proper nouns and make mistakes, but it’s more than good enough to get the general gist of a conversation, even when people are using far-from-ideal laptop mics.
Liang says much of this is down to the company’s own AI engine, which has been tailored to transcribing long conversations instead of the short snippets that the well-known voice assistants are trained to decipher.
“It’s actually a really complicated situation,” said Liang of a multi-speaker online meeting. “A meeting can last for a really long time - it’s a long-form conversation, in contrast to a short question on Alexa, which lasts usually only a few seconds.”
“So we decided to build the technology ourselves... which actually has higher accuracy than the Google API and Microsoft voice API”.
One feature of Otter.ai that you rarely see in other similar services is that the system will often make corrections to the transcription on the fly, as the context of the sentence becomes clearer.
“Sometimes you see Otter go back and correct some previous words [in a live transcript],” said Liang. “This is where the AI is in play, in real time, because when it hears more words it actually re-evaluates the probability of previous words, and if it thinks the probability changes it actually corrects words. This is where it is common-sense aware.”
Liang says the company wants to avoid being tied to any meeting service in particular. “We designed Otter to be agnostic because sometimes you use Google meets, you sometimes use Zoom, maybe Skype or Microsoft Teams or WebEx, so we don't want to limit,” he said. “We want to support all of them.”
I have been a technology writer and editor for more than 20 years. I was assistant editor of The Sunday Times’ technology section, editor of PC Pro magazine and have…