Image credit: read
Summarization is one of the common use cases for various AI models. Multiple tools have created abstracts for articles, PDFs, videos and transcripts. The meeting intelligence tool Read introduces a new feature that trims hour-long meetings into two-minute clips with important pointers.
The company said it uses large-scale language models combined with video analytics to understand the most notable parts of meetings. Read also incorporates participant reactions into the highlight reel. The user navigates to the meeting recording and clicks[ハイライトのみを再生]You can toggle off to see the summarized clip.
Read was co-founded by former Foursquare CEO David Shim in 2021 to collect conference information. Working with Zoom, Google Meet, Microsoft Teams, and Webex, the tool provides analytics such as sentiment and participant engagement scores to provide insight into meeting effectiveness. The company has raised his $10 million seed funding from investors including Madrona Venture Group and PSL Ventures.
Shim said watching an hour-long meeting recording is very boring. He compared watching the reels for two minutes to watching the highlights of the sport his game. He also noted that the company found significant productivity gains for some clients during the testing period.
“During the preview period, agency clients saw an increase in employee productivity of more than 30%. By sharing highlights, we no longer need to set up meetings to talk about our last meeting,” he said in a statement.
In January, Read introduced text-based summaries powered by OpenAI’s GPT model. The Video Highlight feature is an evolution of that. In the demo below, the highlights sometimes get cut off in the middle of the speaker’s sentences when trying to catch the key points. The company said it is working to improve this aspect.
Read also noted that in the coming weeks it will use captions to incorporate text summaries into videos.
Many companies, including Otter and Zoom, are working to provide AI-powered meeting summaries. Most of them use some kind of large-scale language model, but Read’s approach with short TikTok-style video summaries can appeal to those who want to skim through multiple missed meetings.