Show HN: Hurston-1 open source TTS that speak Southern American and Afro-English

Hurston-1 is an open text-to-speech model developed by Destined AI to generate English speech in accents often flattened or poorly represented by mainstream systems. Fine-tuned from Maya1, it supports Southern American and broader American English, as well as Nigerian, West African, East African and Southern African English. Users steer output with plain-language descriptions specifying gender, age, accent, speaking style and pace; each generation creates a voice matching the description rather than cloning a real person. The project, built by a small Atlanta team, frames its work as expanding whose voices are represented in AI and names the model in tribute to writer and anthropologist Zora Neale Hurston.

The model is designed for both standard audio and telephone use. Its default output is 24 kHz, while a phone mode produces 8 kHz G.711 mu-law audio for telephony systems such as Twilio. The team says phone-mode training includes about 10 hours of call-quality speech: 7,600 turns from 250 contributors, alongside recordings from paid, consenting speakers. The model card provides command-line and Python examples, accent and voice-description options, and sample audio. It is released under the Apache 2.0 license. Its stated limitations include approximate rather than exact control over speaking pace, and output that reflects patterns in its training data; it is not intended to reproduce any individual’s voice.