Hurston-1
Text-to-speech for Southern and African accents on phones
AIOpen sourceOpen source (repo not verified)#text-to-speech#accents#telephony
Listed by the ProductPerch team from Hacker News · not launched here yet
What is Hurston-1?
Hurston-1 is an open-source text-to-speech model designed to generate natural English speech with Southern American and various African accents. It was created by Destined AI using consented recordings from contributors across the American South and Africa.
The model includes a narrowband TTS mode built specifically for telephone networks. It outputs 8 kHz G.711 audio formatted for Twilio and other telephony systems, preserving the clarity heard in actual customer service calls.
Users control the output through plain-language voice descriptions specifying gender, age, accent, style, and pace. The system generates new voices that match these prompts without cloning existing individuals.
Hurston-1 features
- Generates Southern American and multiple African English accents
- Outputs 8 kHz G.711 audio optimized for telephone networks
- Controls voice output using plain language descriptions
- Fine-tuned from Maya1 with Apache-2.0 licensing
- Trained on real phone calls and studio recordings
- Runs locally via Python command line interface
Who uses Hurston-1 and for what
- Powering customer service IVR systems with realistic regional accents
- Generating synthetic training data for voice recognition models
- Creating accessible audio content for callers with low bandwidth
- Building telephony applications that require diverse speaker voices
- Testing voice agent responses over standard phone lines
Hurston-1 FAQ
Is Hurston-1 free to use?
Yes, the model is released under an Apache-2.0 license and can be downloaded and run locally at no cost. The website does not mention commercial pricing tiers.
Does Hurston-1 clone real voices?
No, the documentation states that Hurston-1 does not clone voices and was not trained to reproduce any specific real person. Each generation samples a new voice matching the prompt.
What accents does Hurston-1 support?
The model supports Southern American English, Nigerian English, West African English, East African English, and Southern African English, alongside broader American English.
Can I use Hurston-1 for phone calls?
Yes, it includes a narrowband TTS mode that filters audio to the telephone band, resamples it to 8 kHz, and encodes it as G.711 mu-law for telephony systems like Twilio.
How do I control the voice output?
You provide a plain-language description including gender, age, accent, style, and pace. The model generates speech that approximates these parameters.
Comments (0)
Comments and upvotes open when this product launches on ProductPerch.
More AI launches
Generate and edit images in plain language.
Star in the viral Bulin 47 freestyle video from one photo
Cinematic AI video from text, images, and audio
AI sprite animations for indie 2D game builders
Turn one photo into your own State Property meme video
Turn a portrait into a 23-second AI party video
