Hurston-1

专为电话线路打造、由自然语言描述控制的口音英语文本转语音（TTS）模型。

Hurston-1 引入了 NB-TTS（Narrowband TTS，窄带文本转语音）：专为电话线路量身定制的语音合成系统。即便对方是 AI 智能体（Agent），绝大多数客户服务依然发生在电话线上。Hurston-1 的一部分训练数据来自真实贡献者与 AI 语音 Agent 之间的实际通话，其电话模式可以直接输出通话者真实听到的音频：8 kHz G.711 格式音频，开箱即用支持 Twilio 及其他电话通信系统。当然，完整的 24 kHz 音频依然是默认输出。

Hurston-1 是由 Destined AI 基于 Maya1（Apache-2.0 协议）微调打造的模型，旨在呈现绝大多数 TTS 系统处理欠佳的英语口音：包括美国南部口音、尼日利亚口音、西非口音、东非口音、南部非洲口音，以及更广泛的通用美式英语。它是 Destined AI 旗下 ClarityVoice 系列的首个开源模型。

bash pip install torch transformers snac soundfile scipy speechbrain
python hurston.py --model . --check
--accent "Nigerian English accent" --gender Female
--text "Good afternoon. Please take one tablet twice a day after meals." --out hello.wav

python hurston.py --model . --check --phone
--accent "Southern American accent" --gender Male
--text "Thanks for calling, how can I help you today?" --out call.wav

python from hurston import Hurston, describe tts = Hurston('.', check=True)
wav = tts("Well hey there, thanks for calling.", describe('Male', 'adult', 'Southern American accent'))

南部有话要说（The South got something to say）

1995 年的 Source Awards 颁奖礼上，André 3000 面对全场对 OutKast 报以嘘声的观众站在舞台上，说道：“南部有话要说（the South got something to say）。” 我们认为这句话至今依然适用。

Hurston-1 由亚特兰大的一支独立小团队打造。自 2021 年以来，我们一直在有偿且获得明确知情同意的前提下，录制来自美国南部和整个非洲大陆贡献者的声音。在当下大多数 AI 系统中，这些声音依然会被扁平化、同质化为通用的美式英语。没有机构为我们提供融资来做这件事，因此我们完全靠自己完成了研发。这些口音理应在 AI 领域拥有一席之地，如今它们拥有了一个开源模型。

我们将该模型命名为 Zora Neale Hurston。她成长于佛罗里达州的伊顿维尔，接受过专业的人类学家训练，曾踏遍美国南部各地，记录南部黑人真实的说话方式——他们的故事、歌曲和语言表达，而当时几乎没有学术机构认为这些声音值得留存。我们的工作正是对她学术精神的一份小小致敬。该名称仅为表达敬意，并不代表其遗产管理委员会的任何官方背书。

聆听南部之声

以下是几句亚特兰大周边人们日常对话的典型表达。全部为合成语音；每一个声音均根据自然语言描述实时生成。

文本内容	音色描述	试听
"The South got somethin to say."	Male, Southern American accent, clear conversational
"Meet me on Peachtree. No, not that Peachtree, the other one."	Female, Southern American accent, casual conversational
"I'm fixin' to head out, traffic on 285 is already backed up."	Male, Southern American accent, casual conversational
"What part of Atlanta you from?"	Female, Southern American accent, casual conversational

更多涵盖各种口音的音频示例请见 demo/（描述与文本清单位于 demo/demos.csv）。

声音描述规范

该模型基于形如 Gender, age, accent, style, pace（性别, 年龄, 口音, 风格, 语速）的提示词进行训练：

字段	可选值
Gender（性别）	Female, Male
Age（年龄）	adult, middle-aged, older
Accent（口音）	American accent, Southern American accent, Nigerian English accent, Nigerian (Yoruba) English accent, Nigerian (Igbo) English accent, West African English accent, East African English accent, East African (Swahili) English accent, Southern African English accent
Style（风格）	clear conversational voice, casual conversational voice, clear audiobook narration voice
Pace（语速）	natural pacing, measured pacing, slow pacing, fast pacing

语速标签是从训练数据的自然声音中学习到的绝对语速：大致标准为慢速（slow）低于每秒 2 个单词，适中（measured）为 2 到 2.6 个，自然（natural）为 2.6 到 3.4 个，快速（fast）高于 3.4 个。模型对语速标签的遵循是近似的，而非绝对精确，但整体上高度契合底层数据集的分布特征。

每次函数调用都会根据描述抽样生成一个全新的声音。注意：Hurston-1 不具备声音克隆功能，也未经过复现任何真实个体的训练。

专为电话通信打造（NB-TTS）

大多数语音 AI 都在录音室品质下进行演示，但最终却被部署在电话线路上。Hurston-1 则是专门针对其实际应用场景打造的：

• 基于真实电话语料训练。 训练音频中约有 10 小时（来自 250 位贡献者的 7,600 个轮次）属于电话音质语音，源自真实用户与全双工 AI 语音 Agent 的通话，此外还包含录音室风格的朗读和对话音频。
• 电话模式。 --phone（或代码中 phone=True）会将音频输出滤波限制在 300 至 3,400 Hz 的电话频带内，重采样至 8 kHz 并编码为 G.711 mu-law（或通过 --codec alaw 指定为 A-law），这正是电话网络、SIP 协议和 Twilio 所使用的标准格式。这是一种合成后应用的信道效应，因此 --check 质检机制依然会先对全音质输出进行评估。
• 原生流式传输支持。 ulaw_encode() 可直接返回原始的 mu-law 字节流，完美适配 Twilio 媒体流所期望的数据负载（Payload）。

python from hurston import Hurston, describe, ulaw_encode tts = Hurston('.', check=True) call = tts("Hi, I'm calling about my appointment on Tuesday.", describe('Female', 'older', 'Nigerian English accent'), phone=True) payload = ulaw_encode(call)

开发者能用它构建什么

以下是一些切入方向。在 Destined AI 内部，我们使用此类带口音的 TTS 来构建广泛、高质量的测试数据集，这也是我们首要推荐的用法。

• 语音识别（ASR）测试数据。 用 9 种不同口音生成完全相同的句子，检验你的 ASR、字幕生成或转录系统在何处出现崩溃。它非常适合在发版前进行口音覆盖度快速校验或构建回归测试集。
• 语音 Agent 与 IVR 的模拟呼叫者。 将 Hurston-1 与负责编写呼叫者脚本的 LLM（扮演客户、患者、申请人等）结合，在真实用户接入前，通过你的 Agent 运行数百次模拟合成呼叫。开启电话模式，让 Agent 提前听到生产环境中的真实信道音频。这非常有助于发现 Agent 容易听错口音、姓名或数字的盲区。
• 训练数据增强。 在特定口音数据匮乏的场景下，将带口音的合成语音加入到 ASR、关键词检测（Keyword-spotting）或唤醒词训练集中。建议将其与真实语音混合使用，避免仅在纯合成音频上训练。
• 语音提示与播报系统。 电话语音菜单（电话模式的输出可直接导入 IVR 系统）、预约提醒、商场与公共交通播报、App 新用户引导等，用贴近所服务社区真实口音的声音来进行播报。
• 教育与培训。 为英语学习者提供世界各地英语口音的听力练习，或为客户服务、医疗护理及呼叫中心员工提供口音熟悉度培训。
• 无障碍与朗读。 以更具亲和力、更贴近家乡口音的声音朗读文章、表单或通知消息。
• 游戏、动画与影视草稿。 用于背景 NPC 角色、临时配音音轨，以及播客或有声书草稿制作，需明确标明为合成语音。
• 学术研究。 用于语音系统的口音公平性审计、听觉与感知研究，以及覆盖大多数数据集所遗漏口音的基准评测。

目前暂不适用的场景： 跨会话保持完全一致的品牌专属音色（每次调用都会生成新音色；详见局限性）、克隆特定真实人物，或在未经人工试听把关的情况下直接用于安全关键型任务。

评测表现

评测核心关注两个问题：口音听起来像不像？单词发音是否准确？

口音纯正度对比基础版 Maya1。 基础版 Maya1 官方文档记录了 5 种口音（美音、印音、中东音、亚裔美音、英音）。为了公平对比，我们按照 Maya1 官方规范的提示词风格指明各口音，在相同的 10 个句子和 18 个声音设置下与 Hurston-1 进行对比（每个模型评测 180 个音频片段）。使用在真实发音人数据上训练的口音分类器（非洲英语 vs 美式英语）对每组声音进行打分：

口音类型（被判定为非洲英语的概率）	Maya1 基础版	Hurston-1
南部非洲（Southern African）	0.60	0.95
西非（West African）	0.60	0.84
东非（East African）	0.81	0.98
东非斯瓦希里（East African Swahili）	0.81	0.97
尼日利亚（Nigerian）	0.82	0.94
尼日利亚伊博（Nigerian Igbo）	0.82	0.92
尼日利亚约鲁巴（Nigerian Yoruba）	0.87	0.93
美音（American，理论值应偏低）	0.12	0.02
美国南部口音（理论值应偏低）	n/a	0.01

由于该分类器无法区分美国南部口音与通用美式英语，南部口音纯正度主要通过人工试听确认。

清晰度与可懂度。 采用 Whisper large-v3-turbo 对照输入文本测试字错误率（WER），数值经过文本归一化处理（数值越低越好）。

日常句式（与上述口音测试相同的 180 个音频片段，采用 natural pacing 语速，单次生成）：

| Maya1 基础版 | Hurston-1 ---|---|--- 平均 WER | 0.02 | 0.03

高难度复杂句式（30 个未见句式 × 17 个声音设置，共 510 个音频片段）：

| Hurston-1（单次生成） | Hurston-1（启用 --check 自动校验） ---|---|--- 平均 WER | 0.145 | 0.062 WER > 20% 的片段占比 | 17.1% | 8.4%

高难度测试集覆盖了朗读、医疗健康、客户服务、专有名词与数字、高表现力表达及长句等多种场景。基础版 Maya1 之所以非常清晰，是因为它基本上都停留在标准美式英语发音；而 Hurston-1 在真实呈现各种口音的同时，依然保持了几乎同等水平的清晰度。

口音分类（510 片段难句测试集，启用 --check 的 WER）	Hurston-1
美式口音	0.021
美国南部口音	0.070
尼日利亚口音	0.070
尼日利亚（伊博）口音	0.016
尼日利亚（约鲁巴）口音	0.078
西非口音	0.059
东非口音	0.043
东非（斯瓦希里）口音	0.019
南部非洲口音	0.053

人名和数字是最具挑战的类别（启用 --check 时 WER 为 0.17）。其中部分原因在于语音识别器将类似 Adaeze Okafor 等非洲名字拼写错误，而非模型本身读音有误。

语速表现。 在 natural pacing 设置下中位语速为每秒 2.84 词，而基础版 Maya1 为 3.76 词。Hurston-1 的表达更为沉稳清晰；如需更快语速，建议指定 fast pacing。

盲听测试。 测试人员在随机顺序下盲听基础版 Maya1 与 Hurston-1 生成的同一句子（隐去模型标识），并挑选出更贴近真实发音人且符合目标口音的音频。在 54 组对比中，Hurston-1 的胜率高达 78%，基础版 Maya1 仅为 6%；其余判定为持平或均不满意。在非洲口音或美国南部口音上，基础版 Maya1 的胜率为 0%；其仅有的少数胜出全部集中在标准美式英语上。

训练数据

基于 Maya1 采用 LoRA（rank 64）进行微调，训练 2,000 步，包含约 190 小时语音：

数据源	时长	开源许可
Destined AI 口音英语语料库：文本朗读、对话语音以及约 10 小时真实用户与 AI 语音 Agent 通话，均来自获得知情同意的有偿贡献者	约 136 小时（提取自我们 1,800+ 小时的完整语料库）	Destined AI 自有产权
LibriTTS-R（有声书风格，用于保持通用音质基础）	约 46 小时	CC BY 4.0
Mozilla Common Voice 英语非洲口音发音人子集，经过音质过滤	约 10.5 小时	CC0

口音标签来自各贡献者的自报背景，并经过音频口音分类器和人工听音双重验证；对于音频特征与标签不符的样本，均进行了重新标注或剔除。语速标签取自各音频片段实际测得的语速。所有数据贡献者均为成年人。Destined AI 自 2021 年起录制的完整语料库（1,800+ 小时）未随本次发布全部公开；如有商业授权需求请与我们联系。

局限性

• 无法跨调用保持音色一致。 每次调用都会随机抽取一个符合描述的全新音色。长文本在句末会被切分成约 40 词的文本块，因此文本块之间的音色可能会发生漂移。固定命名的专属音色计划在后续版本中推出。
• 专有名词、日期和金额数值是目前表现相对薄弱的内容。在上线重要应用前，请务必启用 --check 选项并进行人工试听审核。
• 泛化的 African English accent 标签效果相对较弱（WER 达 0.26）；建议指定到具体的细分区域口音。
• 仅支持英语。暂无儿童音色。
• 与其他自回归 TTS 模型类似，偶发会出现发音错乱、时长过长或单次生成中途音色突变的情况；开启 --check 可以拦截绝大部分此类异常。
• 输出音频会对所有声音进行统一响度均衡处理，单次生成内部超过 0.5 秒的停顿会被自动缩短。
• 电话模式改变的是信道特性，而非说话风格：模型目前尚不能按需呈现特定电话场景风格（如自然的轮次轮换交替、反馈副语言 backchannels）。详见未来规划。

未来规划

Hurston-1 只是第一步。我们的长期演进方向包括：

• 电话原生 TTS（Telephony-native TTS）。 打造一个真正学会像人类在电话通话中那样“交谈”的模型，而不仅仅是模仿电话线路的物理音效：拥有与朗读风格截然不同的通话风格，具备自然的接话轮次轮转、“嗯哼（mm-hm）”等反馈副语言，以及真实的通话节奏，这些都将从我们录制的真人与全双工 AI Agent 的真实通话语料中汲取学习。
• 音色一致性（Consistent voices）。 为 Agent 和品牌提供跨会话、跨句式完全统一的命名专属音色。

欢迎在 Hugging Face 上关注 Destined AI on Hugging Face，获取下一版本的更新动态。

负责任使用倡议

严禁使用 Hurston-1 伪造他人身份、欺骗听众或从事欺诈活动。在商业部署中，请向听众明确告知该语音属于 AI 合成语音。demo/ 目录下的音频均为合成语音。

致谢与开源协议

• 模型权重：Apache-2.0 协议，衍生自 Maya Research 的 Maya1（Apache-2.0）。
• 音频编解码器：Hubert Siuzdak 开发的 SNAC 24 kHz（MIT 协议），运行时自动下载。电话模式采用 ITU-T G.711（公开通用标准）。
• 可选质检模块：OpenAI 开发的 Whisper large-v3-turbo（MIT 协议）及来自 SpeechBrain 的 ECAPA 说话人嵌入向量（Apache-2.0 协议），运行时自动下载。
• 训练数据：LibriTTS-R（Koizumi et al., 2023; CC BY 4.0）、Mozilla Common Voice（CC0）以及 Destined AI 专有语料库。

引用

bibtex @misc{hurston2026, title = {Hurston-1: Accented English Narrowband Text-to-Speech (NB-TTS) for the Phone}, author = {{Destined AI}}, year = {2026} }

Hurston-1

Accented English text-to-speech built for the phone, steered by a plain-language voice description.

Hurston-1 introduces NB-TTS (Narrowband TTS) : text-to-speech made for the phone line, where most customer service still happens, even when the other side is an AI agent. It learned in part from real phone calls between our contributors and AI voice agents, and its phone mode delivers what a caller actually hears: 8 kHz G.711 audio, ready for Twilio and other telephony systems. Full 24 kHz audio is still the default.

Hurston-1 is a fine-tune of Maya1 (Apache-2.0) built by Destined AI to speak English accents that most TTS systems handle poorly: Southern American, Nigerian, West African, East African, Southern African, alongside broader American English. It is the first open model in Destined AI's ClarityVoice family.

pip install torch transformers snac soundfile scipy speechbrain   
python hurston.py --model . --check \
    --accent "Nigerian English accent" --gender Female \
    --text "Good afternoon. Please take one tablet twice a day after meals." --out hello.wav


python hurston.py --model . --check --phone \
    --accent "Southern American accent" --gender Male \
    --text "Thanks for calling, how can I help you today?" --out call.wav



from hurston import Hurston, describe
tts = Hurston('.', check=True)          
wav = tts("Well hey there, thanks for calling.", describe('Male', 'adult', 'Southern American accent'))

The South got something to say

In 1995 André 3000 stood on stage at the Source Awards, in front of a crowd that booed OutKast, and said "the South got something to say." We think that line still fits.

Hurston-1 was built in Atlanta by a small, independent team. Since 2021 we've been recording consented, paid contributors from the American South and from across Africa. These are voices that most AI systems still flatten into generic American English. Nobody handed us a funding round to do this, so we built it ourselves. These voices belong in AI, and now they have an open model.

We named it for Zora Neale Hurston. She grew up in Eatonville, Florida, trained as an anthropologist, and traveled the South recording how Black Southerners actually talked: their stories, songs and speech, at a time when few institutions thought those voices were worth keeping. Our work is a small echo of hers. The name is a tribute; it does not imply any endorsement by her estate.

Hear the South

A few lines the way people actually talk around Atlanta. All synthetic; every voice is generated from a description.

Line	Voice	Listen
"The South got somethin to say."	Male, Southern American accent, clear conversational
"Meet me on Peachtree. No, not that Peachtree, the other one."	Female, Southern American accent, casual conversational
"I'm fixin' to head out, traffic on 285 is already backed up."	Male, Southern American accent, casual conversational
"What part of Atlanta you from?"	Female, Southern American accent, casual conversational

More samples for every accent are in demo/ (descriptions and text in demo/demos.csv).

Voice descriptions

The model was trained on descriptions of the form Gender, age, accent, style, pace :

Field	Values
Gender	Female, Male
Age	adult, middle-aged, older
Accent	American accent, Southern American accent, Nigerian English accent, Nigerian (Yoruba) English accent, Nigerian (Igbo) English accent, West African English accent, East African English accent, East African (Swahili) English accent, Southern African English accent
Style	clear conversational voice, casual conversational voice, clear audiobook narration voice
Pace	natural pacing, measured pacing, slow pacing, fast pacing

Pace labels are absolute speaking rates learned from the natural voices in training data: roughly slow is under 2 words per second, measured 2 to 2.6, natural 2.6 to 3.4, fast over 3.4. The model follows them approximately, not exactly. It still highly matches the behavior of the underlying dataset.

Each call samples a new voice that fits the description. Note: Hurston-1 does not clone voices and was not trained to reproduce any real person.

Built for the phone (NB-TTS)

Most voice AI is demoed in studio quality and then deployed over a phone line. Hurston-1 is built for where it will actually be heard:

• Trained on real calls. About 10 hours of the training audio (7,600 turns from 250 contributors) is phone-quality speech from real calls with full-duplex AI voice agents, alongside studio-style reading and conversation.

• Phone mode. --phone (or phone=True) filters the output to the 300 to 3,400 Hz telephone band, resamples it to 8 kHz and encodes it as G.711 mu-law (--codec alaw for A-law), the formats phone networks, SIP and Twilio use. It is a channel effect applied after synthesis, so --check still judges the full-quality take.

• Ready to stream. ulaw_encode() returns raw mu-law bytes, the payload Twilio media streams expect.

  from hurston import Hurston, describe, ulaw_encode tts = Hurston('.', check=True) call = tts("Hi, I'm calling about my appointment on Tuesday.", describe('Female', 'older', 'Nigerian English accent'), phone=True) payload = ulaw_encode(call)

What people build with it

Some places to start. At Destined AI we use accented TTS like this to build broad, high-quality test data, and that is still the use we would point to first.

• Test data for speech recognition. Generate the same sentences in nine accents and see where your ASR, captioning or transcription system breaks. It makes a quick accent-coverage check or a regression set before a release.
• Simulated callers for voice agents and IVR. Pair Hurston-1 with an LLM that writes caller scripts (customers, patients, applicants) and run hundreds of synthetic calls through your agent before real people do. Use phone mode so the agent hears what it will hear in production. Good for catching where an agent mishears accents, names or numbers.
• Training data augmentation. Add accented synthetic speech to ASR, keyword-spotting or wake-word training sets where those accents are thin. Mix it with real speech instead of training on synthetic audio alone.
• Voice prompts and announcements. Phone menus (phone mode output drops straight into an IVR), appointment reminders, store and transit announcements, app onboarding, in a voice that sounds like the community you serve.
• Education and training. Listening practice across world Englishes for learners, and accent-familiarity practice for customer service, healthcare and call-center staff.
• Accessibility and read-aloud. Read articles, forms or messages aloud in a voice closer to home.
• Games, animation and media drafts. Background characters, temp voice tracks and podcast or audiobook drafts, labeled as synthetic.
• Research. Accent-fairness audits of speech systems, listening and perception studies, and benchmarks that cover voices most datasets leave out.

Not a fit yet: one consistent brand voice across sessions (each call is a new voice; see Limitations), cloning a specific person, or anything safety-critical without a human listening first.

Evaluation

Two questions: does it sound like the accent, and does it say the words?

Accent strength vs base Maya1. Base Maya1 documents five accents (American, Indian, Middle Eastern, Asian American, British). For a fair comparison we prompted it in its own documented description style, naming each accent, and compared it with Hurston-1 on the same 10 sentences and 18 voices (180 clips per model). An accent classifier (African vs American English, trained on real speakers) scored each voice group:

Accent (probability judged African)	Maya1 base	Hurston-1
Southern African	0.60	0.95
West African	0.60	0.84
East African	0.81	0.98
East African (Swahili)	0.81	0.97
Nigerian	0.82	0.94
Nigerian (Igbo)	0.82	0.92
Nigerian (Yoruba)	0.87	0.93
American (should be low)	0.12	0.02
Southern American (should be low)	n/a	0.01

The classifier cannot tell Southern American from general American English; that accent was checked by listening.

Intelligibility. Whisper large-v3-turbo word error rate (WER) against the input text, numbers normalized (lower is better).

Everyday sentences (the same 180 clips as the accent test above, natural pacing, single take):

| Maya1 base | Hurston-1
---|---|---
Mean WER | 0.02 | 0.03

Harder sentences (30 unseen sentences × 17 voices, 510 clips):

| Hurston-1, single take | Hurston-1, with --check
---|---|---
Mean WER | 0.145 | 0.062
Clips over 20% WER | 17.1% | 8.4%

The harder set covers reading, health, customer service, names and numbers, expressive and long sentences. Base Maya1 is very clear because it mostly stays in American English; Hurston-1 keeps nearly the same clarity while actually doing the accent.

Accent (510-clip set, WER with --check)	Hurston-1
American	0.021
Southern American	0.070
Nigerian	0.070
Nigerian (Igbo)	0.016
Nigerian (Yoruba)	0.078
West African	0.059
East African	0.043
East African (Swahili)	0.019
Southern African	0.053

Names and numbers are the hardest category (WER 0.17 with --check). Part of that is the recognizer misspelling African names such as Adaeze Okafor rather than the model mispronouncing them.

Speaking rate. Median 2.84 words per second with natural pacing, against 3.76 for base Maya1. Hurston-1 speaks a little more deliberately; use fast pacing for quicker delivery.

Blind listening. Testers heard the same sentence from base Maya1 and Hurston-1 in random order, without knowing which was which, and chose the one that sounded more like a real person with the requested accent. Hurston-1 was chosen 78% of the time and base Maya1 6% ; the rest were rated a tie or neither (54 comparisons). Base Maya1 was never chosen for an African or Southern American voice; its few wins were on plain American English.

Training data

LoRA (rank 64) fine-tune of Maya1, 2,000 steps, about 190 hours of speech:

Source	Hours	License
Destined AI accented English corpus: scripted reading, conversational speech and ~10 h of real phone calls with AI voice agents, from consented, paid contributors	~136 (a slice of our 1,800+ hour corpus)	Owned by Destined AI
LibriTTS-R (audiobook style, keeps general quality)	~46	CC BY 4.0
Mozilla Common Voice English, African-accented speakers, quality-filtered	~10.5	CC0

Accent labels come from each contributor's self-reported background, checked against an audio accent classifier and human listening; speakers whose audio and labels disagreed were relabeled or removed. Pace labels come from each clip's measured speaking rate. All contributors were adults. The full Destined AI corpus (1,800+ hours, recorded since 2021) is not included in this release; contact us about licensing it.

Limitations

• No voice consistency across calls. Each call is a fresh voice that fits the description. Long text is split at sentence ends into chunks of about 40 words, so the voice can change between chunks. Consistent named voices are planned for a later release.
• Names, dates and amounts are the weakest content. Use --check and listen before you ship anything important.
• The generic African English accent label is weaker (WER 0.26); use a specific region instead.
• English only. No child voices.
• Like other autoregressive TTS models it occasionally produces a garbled or overlong take, or the voice shifts mid-take; --check catches most of these.
• Output is leveled to the same loudness for every voice, and pauses longer than half a second inside a take are shortened.
• Phone mode changes the channel, not the speaking style: the model does not yet choose call-style delivery (turn-taking, backchannels) on request. See What's next.

What's next

Hurston-1 is the first step. What we are building toward:

• Telephony-native TTS. A model that learns to talk like a person on a call, not just sound like the phone line: a call style distinct from reading style, with natural turn-taking, "mm-hm"s and call pacing, learned from our recordings of real people talking with full-duplex AI agents.
• Consistent voices. The same named voice across every call and every sentence, for agents and brands.

Follow Destined AI on Hugging Face for the next release.

Responsible use

Do not use Hurston-1 to impersonate real people, deceive listeners, or commit fraud. Tell listeners when speech is synthetic. Audio in demo/ is synthetic.

Credits and licenses

• Model weights: Apache-2.0, derived from Maya1 by Maya Research (Apache-2.0).
• Audio codec: SNAC 24 kHz by Hubert Siuzdak (MIT), downloaded at run time. Phone mode uses ITU-T G.711 (public standard).
• Optional checking: Whisper large-v3-turbo by OpenAI (MIT) and ECAPA speaker embeddings from SpeechBrain (Apache-2.0), downloaded at run time.
• Training data: LibriTTS-R (Koizumi et al., 2023; CC BY 4.0), Mozilla Common Voice (CC0), and Destined AI's own corpus.

Citation

@misc{hurston2026,
  title  = {Hurston-1: Accented English Narrowband Text-to-Speech (NB-TTS) for the Phone},
  author = {{Destined AI}},
  year   = {2026},
  url    = {https://destined.ai}
}