Share this guide

Vocal disfluencies, also known as filler words, are sounds or words like “um” and “uh”, commonly found in oral communication. They typically carry no contextual value and are uttered as a speaker transitions between thoughts, filling in the gaps between meaningful words and phrases. As an inescapable part of human communication, they are usually considered a great plague in the arena of public speaking and something to be eradicated at all costs. 

As a result, many customers prefer clean copy transcripts that omit these empty words, but at Deepgram, we know that we serve a diverse customer base with disparate needs. And for some customers–especially those focused on improving the public speaking capabilities of particular end users or tasked with official record keeping–filler words are as important or more important than the words that surround them. These customers require the inclusion of every encountered utterance and place great value on precise, verbatim transcripts. Sample use cases where filler words matter include:

  • Sales enablement

  • Public speaking coaching

  • English language coaching

  • Legal transcription

  • Human Resources transcription

Today, we are happy to announce the release of our new Filler Words feature, capable of transcribing filler words and disfluencies for both pre-recorded and streaming English audio. Filler Words is compatible with existing features like Smart Formatting and Diarization and initially available with our Nova general speech-to-text model, with other model support to be added shortly. Filler Words has no impact on latency or performance and is consistent in its spelling of disfluencies throughout the resulting transcript.

Sample Transcription Output:

Hello, I'm calling about your, uh, home insurance policy. Um, I noticed that your renewal date was coming up and I wanted to reach you to talk about expanding your coverage or possible additions to your plan. So, um, just give me a call back at this number.

To use this new feature, you’ll need to be using our English Nova general model. It’s included for free to hosted and on-prem customers—simply set tier=nova&model=general&filler_words=true to receive verbatim transcripts. 

To learn more, please visit our API Documentation, and you can immediately try out our models and features in our API Playground.

If you have any feedback about this post, or anything else regarding Deepgram, we'd love to hear from you. Please let us know in our GitHub discussions or contact us to talk to one of our product experts for more information today.

Unlock language AI at scale with an API call.

Get conversational intelligence with transcription and understanding on the world's best speech AI platform.

Sign Up FreeBook a Demo