Deepgram today announced a $47 million equity financing round led by Madrona, with Alkeon joining as a new investor. Prior investors also took part in the transaction. Deepgram’s Series B round, in which the company raised a total of $72 million, is now complete. This is the largest known Series B round in the automatic speech recognition area.
Deepgram, which was founded in 2015, specializes on developing specialized voice-recognition solutions for clients such as Spotify, Auth0, and even NASA. Data scientists at the company source, synthesize, classify, and analyse speech data to create speech-recognition models that can grasp brands and jargon, capture a wide range of languages and dialects, and adapt to difficult audio situations. Deepgram, for example, created a model for NASA to transcribe messages between Mission Control and the International Space Station.
Deepgram has big hopes for the future of AI voice recognition. Deepgram uses its audio-first foundation to add nuances of intonation and inflection into sentiment analysis, speaker categorization, summarization, and more, whereas Natural Language Processing has historically followed a text-based approach. To provide conversational context for workplace analytics and expand worldwide, the company has updated its feature set to include subject detection, language detection, and translation.
Deepgram has transcribed over 1 trillion words from over 10,000 years of raw audio data since its inception. NASA, Spotify, and Citi are among the company’s marquee clients. Deepgram’s speech models are the most accurate automatic speech recognition solution in many languages, and the company also provides additional, custom model training as an option to close the gap between human and AI transcription. The new language understanding capabilities of the company add unprecedented context and dimensionality to the future of speech AI technology.