Hurston-1: open 3B narrowband TTS for phone calls, tuned for Southern American and African English accents
Destined AI released Hurston-1, a 3B Apache-2.0 text-to-speech model and LoRA fine-tune of Maya1 on about 190 hours of speech, including consented Southern American and African English recordings, with a phone mode that outputs 8 kHz G.711 audio for Twilio and SIP systems.
Why it matters
Accent realism learned from real regional speech instead of prompt tricks: blind testers chose Hurston-1 over base Maya1 78% of the time for accent realism, and a built-in Whisper check cuts mean word error rate on hard clips from 0.145 to 0.062.
What you could build with it
A small call-centre software vendor could offer regionally natural IVR voices for Southern US and African English markets by running Hurston-1 in its telephony stack, without paying per-minute fees to a closed voice provider.
Does it hold up?
Usable today for phone and IVR prototyping where regional accent realism matters; limits are real — no consistent voice across sessions, English-only, and names, dates and amounts are weakest — so brand-voice production use is premature.
Built with Hurston-1
- A 10-Year-Old CNN Beats GPT-6 at Skin Cancer, and LLM Prices Fall 80%dev.to · Roundup section on Hurston-1 covering the phone mode, the 300-3400 Hz telephone band filter, mu-law encoding for Twilio streams, and accent-coverage test data.
- Hurston-1 open-source TTS targets Southern American and African English accentsAI Briefs · Launch brief noting the Apache-2.0 release, the Atlanta team's consented-speech corpus, and the narrowband phone-call focus.
Learn more
First spotted on article: source.
More AI models
OpenAI publishes 722 AI-generated math manuscripts from an unreleased modelOpenAI released 722 mathematical manuscripts in 372 result families to GitHub under Apache-2.0, produced by an…models · JEV 0.74Tencent Hunyuan Hy3: open 295B MoE with 21B active under Apache 2.0Tencent open-sourced Hy3, a 295B-parameter mixture-of-experts model with only 21B parameters active per token, under…models · JEV 0.73NVIDIA Isaac Lab 3.0 GA + Isaac OS open runtime + Agile ONE SN humanoidMajor Isaac platform expansion at GTC Berlin Oct 21: Isaac Lab 3.0 is generally available as the open-source robot…models · JEV 0.73StepFun Step-5-Preview — open BF16 weights releasedStepFun published the full BF16 weights for its flagship Step-5-Preview agentic model (600B-parameter sparse MoE, 27B…models · JEV 0.73NVIDIA Nemotron 3.5 Retriever collection: Embed 8B/1B, Rerank 8B, Parse 631MNVIDIA's Oct 21 collection drop for agentic retrieval: Nemotron-3.5-Embed in 8B BF16, 1B BF16 and 1B NVFP4…models · JEV 0.72Kimi K3.1 Pro: Moonshot's 2.8T open-weight flagship officially launchesMoonshot AI officially launched Kimi K3.1 Pro on Oct 23, a mixture-of-experts model with 2.8 trillion total / 104…models · JEV 0.72
Get the week's best AI launches, plus 3 ideas worth building
One email every Saturday. Ranked by traction, not hype. Free.