huggingface
An openly available speech synthesis model that creates speech in various English accents and outputs audio for telephone networks.
SITE CAPTUREBased on information published on the service's website. Selection and review policy
WHAT IT DOES
What does it do?
Hurston-1 synthesizes Southern American and various African English accents from natural language voice descriptions. Its default output is 24kHz, and telephone mode produces 8kHz G.711 audio. It can generate the same sentence in multiple accents to check for weaknesses in speech recognition systems.
KEY FEATURES
Key features
- Generates English speech from natural language voice descriptions.
- Supports Southern American and various African English accents.
- Outputs 8kHz G.711 audio in telephone mode.
- Returns mu-law bytes for use with Twilio Media Streams.
- Generates the same sentence in multiple accents to test speech recognition performance.