Try our FREE Self-Assessment. Which speech-to-text provider is right for you? 🤔

Deepgram Changelog

Stay up to date with enhancements to our AI Speech Platform and ecosystem

SEARCH
Five new languages and three new use case speech models

We are announcing five new languages and three new use case speech models.  In addition, we have improved our Hindi and Spanish language models and our Conversational AI and Meetings models.  These improvements mean better…

Updates to our Speech Recognition API

Deepgram has released an update to our Speech Recognition API. Improvements affect the following areas: Keyword Boosting for Pre-Recorded Transcription We have increased the accuracy of out-of-vocabulary (OOV) keyword boosting by reducing false positives when…

Now in General Availability: Utterance Formatting

utterances=true (optional parameter utt_split=number) Our new utterances feature segments speech into meaningful semantic units, which allows the chosen model to interact more naturally and effectively with speakers’ spontaneous speech patterns. For example, when humans speak…

New from Deepgram Labs: Hindi Support

language=hi Deepgram now supports automatic transcription for Hindi with the general model (model=general). The new language has been trained on a variety of data ranging from basic conversations to domain-specific call center audio. To learn…

Improved Spanish Support

language=es new version = 2021-05-19.0 previous version = 2021-02-10.0 We have done additional training to improve the accuracy and expand the vocabulary of our Spanish language support. To use the latest Spanish support, pass the…

Updated Phonecall Model (English US)

model=phonecall version=2021-05-19.0 We have done additional training to our base Phonecall Model for US English to improve accuracy and expand the model vocabulary.  We’ve also incorporated our latest improvements to punctuation (punctuate=true). To use the…

Updated General Model (English US)

model=general version=2021-05-19.0 We have done additional training to our main General Model for US English to improve accuracy and expand the model vocabulary.  We’ve also incorporated our latest improvements to punctuation (punctuate=true). The general model…

Updated Meeting Model (English US)

model=meeting version=2021-05-19.0 We have done additional training to our base Meeting Model for US English to improve accuracy and expand the model vocabulary.  We’ve also incorporated our latest improvements to punctuation (punctuate=true). To use the…

Protected: Invite-Only Beta: Support for Out of Vocabulary Keyword Boosting

There is no excerpt because this is a protected post.

New version of self-hosted Speech Engine for On-Prem Customers

Deepgram released a new version of the self-hosted Speech Engine. Current on-prem customers may pull this version from Docker Hub as deepgram/onprem-engine:3.13.2 or deepgram/onprem-engine:latest. Improvements include: Resolution of CVE-2021-3185 in a supporting library Reliability improvements for real-time…

Improved Turkish Support

language=tr version =2021-04-19.0 We have done additional training to improve the accuracy and expand the vocabulary of our Turkish language support. To use the latest Turkish support, pass the language=tr as an API parameter in…

Open Beta: Redaction – Support for Social Security Numbers

redact=ssn We’ve added beta support for redaction of social security numbers from automatic transcripts. To test the feature, send your audio to beta.deepgram.com and set redact to true. To learn more about the feature and participate…

Open Beta: Utterance Formatting

utterances=true Our new utterances feature segments speech into meaningful semantic units, which allows the chosen model to interact more naturally and effectively with speakers’ spontaneous speech patterns. For example, when humans speak to each other…

Keyword Boosting Effect

On February 17, we released our improved keyword boosting feature which can be used to boost or suppress recognition of certain keywords. We have determined that some words that are unfamiliar to the speech model…

Improved Error Messages for Realtime Streaming

Deepgram has released a change to how we report errors during Realtime Streaming. Now, we will return a WebSocket Close frame and the body of the Close frame will indicate the reason for closing using…

PAGE 1 OF 2

Apply Now

Receive up to $100,000 to use over 12 months.

Become a Partner

When you become a partner you’re in good company.

Talk to Customer Success