Speech Dataset

African American Vernacular Dataset

Afro-Amerikaanse volkstaal datastel

High-Quality African American Vernacular Call-Center, and Podcast Dataset for AI & Speech Models

Contact Us
African American Vernacular Dataset

Overview

  • LanguageAfrican American Vernacular
  • Dataset TypesCall-Center, Podcast
  • CountryAfrican American

Description

Synthetic telephonic conversations between an agent and a customer range from 5 to 15 minutes in duration, while licensable public domain audio or video files such as interviews or podcasts with 1 to 5 participants typically range from 15 to 60 minutes.

Dataset Highlights

  • LanguageAfrican American Vernacular
  • Sample Rate16 kHz / 8 kHz
  • Total Hours310:05:53
  • Speakers1,584+
  • ChannelsDual / Mono
  • Duration Range5 - 60 mins

Use Cases

  • ASR (Automatic Speech Recognition)
  • Virtual Assistants & Chatbots
  • Conversational AI
  • Speech Analytics
  • TTS (Text-to-Speech)
  • Language Modelling

Sample Dataset Specs

Dataset TypeCall CenterMedia Data
Sampling Rate8 kHz16 kHz
Speakers2 SpeakersMultiple Speakers
ChannelDualMono
Total Hours155:40:47154:25:06
Total # of Speakers1,284300

Key Benefits

  • Diverse & RepresentativeBroad speaker coverage for inclusive, real-world AI.
  • High-Quality AudioClean, balanced recordings for reliable model performance.
  • Ready for AI/MLStructured and annotated for seamless integration.
  • Scalable & FlexibleMultiple durations, speakers, and domains to fit your needs.

Featured Clients

Trusted by leading AI teams and enterprises worldwide.

  • Microsoft
  • Amazon Web Services
  • Google

Can’t find what you are looking for?

New off-the-shelf datasets are being collected across all data types

Contact us now to let go of your audio/speech training data collection worries

  • This field is for validation purposes and should be left unchanged.
  • By registering, I agree with Shaip Privacy Policy and Terms of Service and provide my consent to receive B2B marketing communication from Shaip.