Otter AI Chat brings new collaboration and insight to real-time voice transcription

Ever wanted to ask questions about a meeting while it’s still in progress? Otter AI Chat, a new capability from AI-powered voice transcription provider Otter, has now got you covered.

The new Otter AI chat functionality being announced today by the San Francisco startup gives users the power of generative AI as an integrated component of voice transcription. Otter has been busy expanding its voice transcription service in recent years, adding integration with popular conferencing technologies including Zoom and support for Microsoft Outlook. In February of this year the company expanded its OtterPilot AI functionality, bringing new automations to its voice transcription service. The company claims that its AI-powered service transcribes over one million spoken words every minute.

>>Follow VentureBeat’s ongoing generative AI coverage<<

Now Otter is taking its AI a step further than just transcribing and providing a post-meeting summary. With the new Otter Chat capability, an integrated AI chatbot inside the user console enables users to ask questions about a meeting in progress.

A user could, for example, find out if a certain topic has already been discussed. Users can even ask about the sentiment of the meeting or of a specific speaker, and ask the AI to generate content based on the meeting, such as follow-ups and action items.

It’s an approach in many respects inspired by the success of OpenAI’s ChatGPT.

“ChatGPT is such a sensation — it shows that AI can understand human questions and can generate interesting answers,” Sam Liang, founder and CEO of Otter, told VentureBeat. “However, ChatGPT is mostly based on public knowledge; for corporate workers, their meetings are internal and … happening in real time.”

Otter has built its own type of LLM for Otter AI Chat

While Otter AI Chat is inspired by ChatGPT, it is not using the OpenAI technology. Rather Liang emphasized that his team has developed its own purpose-built AI technology to enable the new service.

Liang explained that OpenAI’s ChatGPT is based on a large language model (LLM) trained on public data that consists mostly of written content. ChatGPT provides a conversational interface between the human being and the AI.

Otter AI Chat is taking a somewhat different approach.

“We’re building the largest spoken large language model based on mostly verbal data,” Liang said. “The [spoken] verbal data has a lot of differences compared to written documents.” 

With spoken verbal data, there is more interaction and speakers are often less formal and more casual. There is also the ability to better understand and derive sentiment from the tone of a speaker. Otter has been able to train on over a billion meetings the company has transcribed, while keeping all user information confidential.

Otter building a collaborative chat for AI

Liang said that rather than being just a human-to-bot interface, the Otter AI Chat interface enables what he referred to as “collaborative chat,” where all the people in a meeting can collaborate both with each other and the AI chatbot in real time.

“AI effectively joins human conversations,” Liang said.

As the model was trained on the way humans actually speak, Liang said that Otter AI Chat will also be able to better respond with human-style speech, which will help to make the information more useful and engaging. 

The Otter AI Chat capability is part of the company’s overall aim to use AI to make meetings more useful and effective for participants. With the initial launch, Liang said that the new generative AI chatbot is being made available as a text interface that can be accessed inside of a multi-speaker meeting. The plan for the future is to make the AI chat available via a voice interface as well.

“We’ll make it available through voice as well so that effectively Otter can join your speaking session and you can ask Otter any question with voice and Otter can answer questions with voice as well,” he said.

>>Don’t miss our special issue: Building the foundation for customer data quality.<<