Deepgram: The Google of Audio

We all have used Google to search for anything and everything under the sun. Google’s search engine is so powerful that it can sense what you are intending to search for, even if you can’t phrase it properly as a search query. Now Google is only known for its prowess in search word phrases and to an extent, pictures. But what if the search query needs to be in the form of an audio? Deepgram lets you do just that.

Deepgram, an AI audio startup, has developed a neural network based artificial intelligence software. Using this software, it can search through audio files just like Google does with word queries. The co-founder of Deepgram, Noah Shutty had this idea when he found it hard to search for specific audio life-logs which he had recorded over the years. Using this new software, people can search and categorise audio files just like they do with word documents.

Business streams for Deepgram

Searching for audio recordings is something that is vital for a variety of different fields. Journalists, for instance, record thousands of hours worth of interviews at work. Traditionally, they are forced to listen to the whole tape to understand it completely and to create insights. Deepgram could effectively solve this issue by automating the process. It can produce actionable insights on its own and thus save hundreds of man hours for the journalists and to media houses.

Apart from this, audio recordings are an integral part of customer service divisions of all major corporations. As far as the workflow goes, transcribing audio content is a tedious process. Also, once it is transcribed it needs to be fed in as search queries, making it extremely time-consuming and expensive. By automating this, Deepgram offers a viable and cheaper alternative to these firms.

Process behind the innovation

From the user end perspective, it is just about feeding the Deepgram software with the audio file, or even an URL link to the file. Once this is done, Deepgram processes the speech and stores it as a ‘deep representation index’. This basically is indexing files based on phonetics instead of the actual words.

This is innovative because it lets the users search the file by the way they sound, instead of how they are spelt. So even if the words are misspelt, the system can identify it properly. The final solution is given through an easy-to-use API through which actionable insights can be gathered.

Future of audio search through Deepgram

The company now has over 1000 registered users. It charges clients based on the amount of data that they process through its system. Recently Deepgram had given its services for free to journalists covering the US 2016 elections. This was to get the word out on the street about their technology and about how it can change the way transcribing is done from now on.

Deepgram has recently closed a funding round worth $1.8 million, led by Metamorphic Ventures and Y Combinator. It plans to scale up and commercialise the tool in the enterprise market in the immediate future.