Skip to main content

Otter's free app brings voice transcripts into the AI age

a close up of a computer: An iPhone shows a transcription in the Otter app © Provided by CNET An iPhone shows a transcription in the Otter app If you don't hate transcribing, it's probably just because you don't have to do it very much. 

Otter, a new, free mobile app from a team of vets from Google and speech-recognition company Nuance, aims to make voice transcriptions become as easy and accessible as typing into a Google Doc. 

Voice is an obsession of tech giants right now. But companies like Amazon, Google and Apple are mostly zeroed in on voice-command assistants like Alexa or Siri, premised on the bet that voice interaction will become the next stage of computing. Getting less attention, though, are voice transcriptions -- tech that takes a recording of people talking and turns it into text -- even though that kind of technology could be transformative for people across different needs and professions.  

Most transcription apps or services fall into two buckets. If they're free, they're not often accurate. And those that give you clean transcriptions are usually expensive. That's because voice-to-text is technologically tricky, and most services that provide you with accurate transcripts usually need a human to review the recording. 

Otter, which debuted this week at Mobile World Congress, aims to make transcripts not only free and accurate but also smart. 

This is a perfect time," said Sam Liang, CEO and founder of AISense, the company behind Otter. "With AI tech and deep learning in the last few years, the accuracy for speech recognition has improved dramatically. A few years ago, this system wouldn't be usable."

To use Otter, you tap a microphone icon to begin a recording, and almost immediately, a live raw transcript of what you're saying begins to unspool in front of you. It's after that recording ends that Otter's artificial intelligence can really do its work. After processing, the cleaned-up transcript separates speakers as they take turns talking. As part of your sign-up process, Otter takes a "voiceprint" of you by asking you to read a five-paragraph statement so it can learn your voice and specifically identify you next to the passages you spoke. 

Otter's transcripts also are searchable, not only within one transcript but across all your stored recordings. It automatically generates keywords you can tap to search. You can form teams in Otter, and content can be shared within the Otter app with individuals or team members. You can also send anyone a link to a transcript viewable on the web. 

In one neat feature, the text and audio are synchronized when you play back the recording, so if you tap on any part of the transcript, the audio will jump right to that place for quick accuracy checks. During an audio playback, the app also highlights each word as it's spoken. 

It's not a perfect transcription. Punctuation is sometimes out of whack. It misidentifies some words -- I told Otter, "Hi, I'm Joan" and my readout picked up my name as John, for example. And it's less successful in a crowded environment or during cross-talk. And the app, which was released publicly Monday, is somewhat buggy in its early days. Once when I opened it, all my recordings and their transcripts from the previous day were missing. Restarting the app restored them, but you may notice other bugs. 

And it lacks some tantalizing capabilities, like importing a previously recorded conversation. The app and its baseline transcription service is and will remain free, but the company plans to add a subscription tier later that unlocks extra utilities, along the same lines as Evernote's business model. 

Its privacy relationship with your materials is similar to that of an Evernote, too. Otter is confidential and encrypted. And the company isn't building an ad-based business with Otter, so it's not interested in peeking into your materials so it can create a profile that will target ads to you, the company said. 

Otter is available for devices running Google's Android system and on Apple mobile products.

Popular posts from this blog

R. Tiglao Exposed LTO Records Showing Aquino Never Bought or Sold a Porsche

Veteran columnist Rigoberto Tiglao exposed the Land Transportation Office (LTO) records of former President Benigno Aquino III showing that he never bought or sold an expensive Porsche car. The LTO records proved that the former President did not sold his Porsche 911 Carrera car which he claimed he bought for P5 million. The controversial Porsche car of the former President made headlines just months into his presidency but he explained that he bought the luxury car with the proceeds he got when he sold his BMW. Because of the furor from such display of opulence, Pres. Aquino claimed to have sold it six months later for exactly the same price. According to Tiglao during that time he asked through his column the LTO to release the car's deed of sale and registration to prove that it was not a gift from a Chinese-Filipino tycoon as rumored by some individuals critical to the President. The only possible way to discover whether the Porsche luxury was indeed sold was through ...

You can pay at a restaurant by smiling at a camera

© Provided by Engadget As easy as it is to make purchases in the era of tap-to-pay services , it's about to get easier still. Alipay (which handles purchases for Chinese shopping giant Alibaba) has launched what it says is the first payment system that uses facial recognition to complete the sale. If you visit one of KFC's KPRO restaurants in Hangzhou, China, you can pay for your panini or salad by smiling at a camera-equipped kiosk -- you need to verify the purchase on your phone, but you don't have to punch in digits or bring your phone up to an NFC reader. The system (Smile to Pay) is purportedly resistant to spoofing with photos and other tricks. It relies on both depth-sensing cameras and a "likeness detection algorithm" to make sure it's really you. Reportedly, the technology is good enough that it can accurately identify people even when they're disguising themselves through makeup or wigs. You shouldn't have to worry about someone buying ...

Tesla cloud account hacked to mine cryptocurrency

© Provided by The Hill An unidentified outside hacker infiltrated Tesla's Amazon cloud account and used its systems to quietly mine for cryptocurrencies, a cybersecurity firm announced Tuesday. The hack also potentially exposed the electric car company's data. Researchers for RedLock found that Tesla's credentials on an IT administrative console were not password protected. They made the discovery while trying to track down which organizations had left their Amazon Web Services (AWS) credentials openly exposed on the internet last month. The hackers quietly hijacked the console and began running scripts to generate virtual currency like bitcoin, the latest in a series of "cryptojacking" attacks. The researchers also found the hackers used "sophisticated evasion measures" to go undetected. A spokesperson for Tesla said the company learned about the breach in a company-sanctioned bug bounty program that pays outside hackers to discover vulnerabilitie...