Destined AI's open-source (Apache-2.0) accented-English text-to-speech model — it voices Southern American and African English accents from a plain-language description, with a telephone-ready phone mode.
Last verified · 2026-10-25 · by Moe Ameen
Hurston-1 (`destined/hurston-1`) is an open, Apache-2.0 text-to-speech model from Destined AI, released as the first model in its ClarityVoice family. The model card describes it as accented English "built for the phone, steered by a plain-language voice description." It is a LoRA fine-tune of the Maya1 base model, trained on roughly 190 hours of speech and shipped as a 3B-parameter BF16 checkpoint on Hugging Face (the repo is public but gated behind an access accept).
The distinctive idea is that you do not pick a preset voice or clone a person — you describe one. Each call takes a short brief (gender, age, accent, speaking style, pace) and the model synthesizes a fresh voice to match. Its headline accents are Southern American English and a set of African English variants (Nigerian, including Yoruba- and Igbo-influenced, plus West African, East African, and Southern African), with General American available too. It is English-only and ships no child voices. Default output is 24 kHz through the SNAC codec, and a dedicated phone mode emits 8 kHz G.711 telephone-band audio for IVR and voice-agent work.
You run it from a Python CLI or library. An optional `--check` pass uses Whisper and speaker embeddings to re-synthesize bad takes, which the card says cuts word error rate on harder sentences from about 0.145 to 0.062. Destined AI reports its own evaluations — a blind-listening preference of 78% over the base Maya1, and about 0.03 WER on everyday sentences — and is candid about limits: no voice consistency across separate calls yet, long text split into roughly 40-word chunks between which the voice can drift, and names, dates, and amounts as the weakest content. No official release date is stated beyond a 2026 citation, so treat the specs and figures as a snapshot of the model card.
Think of Hurston-1 as the voice actor and [Kompozy](/) as the studio and the distributor. Hurston-1 is superb at one thing — producing an authentic Southern American or African English voice track from a script — and completely uninterested in what happens next: it hands you a WAV and stops. A single narration is not a channel, and that is precisely the gap Kompozy fills, turning one scripted voiceover into a week of on-brand, captioned, scheduled content.
The connective tissue is Kompozy's [Clipped Shorts](/glossary/clipped-short): once you have laid a Hurston-1 narration over footage or a still to make a video, upload that clip and Kompozy cuts it vertical and burns in word-synced captions — the step that makes an accented voiceover work on feeds that autoplay muted. From the same script, Kompozy then generates the formats Hurston-1 can't touch: a [Carousel](/glossary/hyperframes) built on pixel-exact HyperFrames, quote graphics of the strongest line, per-platform text posts, a blog, and a newsletter, all held to your [Persona Brief](/glossary/persona-brief) so the written copy matches the spoken voice. Kompozy's own [Persona Shorts](/glossary/persona-shorts) add a talking-head segment between narration beats, and [Autopilot](/glossary/autopilot) schedules and publishes the whole set across the eight social platforms plus blog and email behind a per-post review. Hurston-1 gives the words a regional voice; Kompozy turns that one narration into a published grid.
Hurston-1 (destined/hurston-1) is Destined AI's first open ClarityVoice model — an Apache-2.0, 3B-parameter accented-English text-to-speech model, fine-tuned from Maya1. You describe a voice in plain language (gender, age, accent, style, pace) and it synthesizes a new voice, with Southern American and African English accents as its focus.
Yes — it is released under Apache-2.0 with open weights on Hugging Face, though the repo is gated behind an access accept. There is no license fee; your cost is the hardware and Python setup to run it, plus extra dependencies for the optional --check quality pass.
Southern American English, General American, and African English variants (Nigerian, including Yoruba- and Igbo-influenced, plus West African, East African, and Southern African). It is English-only, no child voices. Default output is 24 kHz audio, with a dedicated 8 kHz G.711 phone mode for IVR and voice agents.
Kompozy is the layer around the voice. Draft the script in your Persona Brief voice in Kompozy, narrate it with Hurston-1, then upload the resulting clip to Clipped Shorts for captioned vertical cuts and repurpose the script into carousels, quote cards, text, a blog, and a newsletter — scheduled and published across nine destinations: eight social platforms plus blog and email.