Nobody really enjoys taking notes, but it is one of those unavoidable facets of life for many people in business and academia. During phone calls, specifically, it can be hard to focus on listening to what’s being said while ensuring you accurately jot down what people are saying in real time.
But what if there was a way for you to record a call through your mobile phone, and have a full transcription of the discussion delivered to you within minutes? That’s exactly what San Francisco-based Tetra is setting out to enable with its AI-powered iPhone app that not only records your calls, but also converts the conversations into written form using deep learning and natural language processing (NLP).
Fresh from $1.5 million in seed funding, Y Combinator alum Tetra has announced a notable update to its mobile app. So far, Tetra has only worked with outbound calls, but now subscribers will be able to enjoy the full benefits of Tetra for incoming calls too.
By way of a quick recap, Tetra is basically a VoIP app that works similar to Google Voice insofar as it allocates you a dedicated Tetra number which must be used for all outgoing/incoming calls. Once the call is complete, Tetra will then spend a short period of time generating the notes.
Above: Recording then processing notes
Based on our brief tests, Tetra was able to accurately record and transcribe the audio, though we didn’t really put the app through the wringers in terms of throwing complex names, abbreviations, or colloquialisms into the mix. For normal day-to-day chat, though, it seems to work just fine.
Additional smarts include the ability to search for keywords across all conversations, and then re-listen to a specific piece of audio by tapping on keywords. Users can help guide Tetra’s AI to create call summaries by manually tagging key moments during a call, while they can share notes with colleagues who weren’t present, which can be accessed through the Tetra mobile or web app.
Above: Tetra: Notes and search
In terms of pricing, well, everyone can get 60 free minutes per month as part of a trial. Then, you’ll have to sign up to the “Plus,” Pro,” or “Business” plans which offer varying amounts of call-time per month, ranging from $9 to $99.
Above: Pricing
Then there is the legal / ethical angle to consider. By default, Tetra automatically tells the people on the other end of the call that they are being recorded, however it’s possible for the Tetra subscriber to disable this announcement on the proviso that you “stay compliant with local law or get recording consent yourself,” according to Tetra.
AI for good
Though there is an almost palpable fear regarding what AI will mean for the future of humans, being able to automate boring transcriptions is undoubtedly one benefit that most people will welcome with open arms.
That said, companies who are concerned about data confidentially will likely have some reservations over eavesdropping bots listening in on their calls, while the possibility that some content could be accessible to humans will also be of concern. According to Tetra, it uses “industry standard” encryption and AWS servers to store data. It also said that “no humans are ever involved” in the creation of the transcripts.
Of course, automated transcription isn’t a new concept. Last year, Zoom announced it was adding automatic transcription smarts to its video conferencing platform, while emerging startups such as Clark.ai and Scopist are setting out to integrate with popular third-party conference calling software to automate note-taking. A wearable voice recorder called Senstone also raised north of $300,000 through Kickstarter last year to make it easy for you to convert real-world audio into written notes.
There are existing players in the iPhone app space too, that record calls and transcribe the audio. However, Tetra claims higher levels of transcription accuracy without any human interference, within minutes.
It’s clear there is big demand for automated transcriptions. While AI transcription won’t yet be perfect, and it will likely struggle with more technical and nuanced language, it’s already “good enough” for many use-cases. For now though, Tetra only officially works with U.S. English, though it didn’t seem to have any troubles with my brand of U.K. English, which will actually be officially supported “soon.” Other languages which will soon be supported include Russian, Spanish, Hindi, Japanese, German, Czech, and Polish.
A spokesperson also confirmed to VentureBeat that an Android app is coming within the next few months.
Notable omissions in Tetra’s armory include integrations with third-party tools such as Slack, Skype, and GoToMeeting, though the company’s FAQ answer on whether it works with such communication platforms is “not yet,” which suggests it is open to it.
Still, it’s very early days for Tetra. The company only launched its app in October, and has a team of just five. That is one of the benefits of using off-the-shelf APIs for its AI, as it means it doesn’t have to reinvent the wheel and it can streamline its operations as it seeks growth.
No comments:
Post a Comment