--- title: Show HN: Hurston-1 open source TTS that speak Southern... description: Hurston-1 is an open text-to-speech model fine-tuned from Maya1 to handle English accents most TTS systems get wrong, including Southern American, Nigerian,... canonical: https://daily.dev/posts/show-hn-hurston-1-open-source-tts-that-speak-southern-american-and-afro-english-mccjqt4zw twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Show HN: Hurston-1 open source TTS that speak Southern American and Afro-English | daily.dev og:description: Hurston-1 is an open text-to-speech model fine-tuned from Maya1 to handle English accents most TTS systems get wrong, including Southern American, Nigerian,... og:url: https://daily.dev/posts/show-hn-hurston-1-open-source-tts-that-speak-southern-american-and-afro-english-mccjqt4zw og:image: https://api.daily.dev/og/posts/MCCjQt4zW.png og:image:alt: Show HN: Hurston-1 open source TTS that speak Southern American and Afro-English og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Show HN: Hurston-1 open source TTS that speak Southern American and Afro-English **[Hacker News](https://daily.dev/sources/hn)** · 11 min read · 0 upvotes · 0 comments ## Summary Hurston-1 is an open text-to-speech model fine-tuned from Maya1 to handle English accents most TTS systems get wrong, including Southern American, Nigerian, West African, East African and Southern African. It introduces NB-TTS, a phone mode that outputs 8 kHz G.711 audio ready for Twilio and telephony systems. The model is a LoRA rank-64 fine-tune trained on about 190 hours of speech, including roughly 10 hours from real phone calls with AI voice agents. In blind listening tests it was preferred over base Maya1 78% of the time versus 6%. Weights are released under Apache-2.0. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### How do I generate 8kHz G.711 audio for Twilio from a text-to-speech model? Hurston-1, a fine-tune of Maya1, has a built-in phone mode that filters output to the 300-3,400 Hz telephone band, resamples to 8 kHz, and encodes as G.711 mu-law (or A-law). Running the model with the --phone flag or phone=True produces this output, and ulaw_encode() returns raw mu-law bytes ready for a Twilio media stream. _Developers wiring TTS into Twilio call flows can track tools like this on daily.dev._ ### What open-source text-to-speech model handles African and Southern American English accents well? Hurston-1, an Apache-2.0 fine-tune of Maya1, targets Southern American, Nigerian, West African, East African and Southern African English accents that most TTS systems handle poorly. In blind listening tests comparing it to base Maya1, Hurston-1 was preferred 78% of the time versus 6% for Maya1, which was never chosen for an African or Southern American voice. _Developers evaluating accented speech models for inclusive voice products can follow releases like this on daily.dev._ ### How much training data and compute does it take to fine-tune a TTS model like Maya1 for new accents? Hurston-1 was produced with a LoRA fine-tune at rank 64, run for 2,000 steps, using about 190 hours of speech. Roughly 136 hours came from Destined AI's own accented English corpus, including about 10 hours of real phone calls with AI voice agents, supplemented by quality-filtered Mozilla Common Voice recordings from African-accented speakers. _Teams planning their own LoRA fine-tunes can benchmark data and step counts against examples like this on daily.dev._ ## Community take How the wider developer community reacted, aggregated from 1 discussion and 44 comments across hackernews (as of 2026-10-26). **TL;DR:** The discussion mostly debated the confusing title and terminology (especially around 'Afro-English' and AAVE) while praising the underlying effort to better represent underserved American Southern and African-accented English in TTS. **Sentiment:** 35% positive · 50% mixed · 15% skeptical **The case for** - Several commenters said Southern US and African-region accents are poorly represented in existing speech models, making this a valuable gap-filler. - The author's active, detailed responses explaining methodology and future plans were well received. **The pushback** - The title's phrasing ('Southern American', 'Afro-English') caused real confusion, with some initially reading it as referring to South America. - Terminology like 'Regional African English' and 'Afro-English' was criticized as imprecise or confusing when applied to Black Americans. - One commenter questioned whether the African-accent training data adequately captures the continent's huge linguistic diversity. - The HuggingFace demo/playback didn't work and the model weights appeared gated/restricted despite the Apache-2.0 claim. **By community** - hackernews (mixed): A mix of praise for addressing underserved accents, confusion/criticism over the title and terminology, and practical complaints about access to the demo and model. **Hottest debate:** Whether terms like 'Afro-English' and 'Regional African English' appropriately or confusingly describe Black American versus continental African accents. **Open questions** - How well does the African-accent training data account for the linguistic diversity within the continent (e.g., different native-language influences on English accents)? - Will there be a working public demo, and is full open access to the weights actually available given the gated HuggingFace page? **Highlights** > AAVE was an awful term, but "Regional African English" and "Afro-English" is even worse if you're trying to talk about black Americans. Are you talking about the English of Africans in different parts of Africa? If so, are you including the voices of white and Mediterranean Africans? I'm confused. As a black American with roots (going back a couple centuries) in the Ark-La-Tex, I would think our dialects would be covered under "Southern American," which makes me suspect that you were referring to me with all of those references to another continent. > — [pessimizer on hackernews · 1 comments](https://news.ycombinator.com/item?id=50236915) > Wow this is really cool. AAVE and Southern lexes have been underserved linguistically for a long time. That said, my hunch is your Africa speicifc datasets might lossy, as entire regions of Africa are very linguistically diverse and someone who speaks Bambara and French will affect a different English accent than someone who speaks Maninka and English. > — [alephnerd on hackernews · 1 comments](https://news.ycombinator.com/item?id=50235445) > congrats, this is a very exciting release. I cannot speak to the other accents, but my experience has been that American south accents are really poorly represented in mainstream Speach-To-Text models. If your STT model is pointed at end users, this can cause quite a quality drift when looking at eval vs prod performance depending on who your user base is. I am not sure what terms you have collected the speach data on, but I am sure there are lots of people working with voice models that would be very interested in this as an eval set. > — [islewis on hackernews · 1 comments](https://news.ycombinator.com/item?id=50236217) > Trying to access the wave files directly reveals: Access to model destined/hurston-1 is restricted and you are not in the authorized list. Visit https://huggingface.co/destined/hurston-1 to ask for access. > — [zeroq on hackernews](https://news.ycombinator.com/item?id=50237336) > Yes, exactly. This version focuses on giving the ability to consistently represent those voices well. The future versions will be even more fine-grained where you can scope it even more detailed if you wanted. For example, right now if you say you want a Southern American accent, you might get a voice that is a Black person from Atlanta or Rural Georgia in future versions you will be able to specify Atlanta or rural, etc.. same with the regional African accents. > — [Gcarter08 on hackernews](https://news.ycombinator.com/item?id=50237180) **Source threads** - [hackernews](https://news.ycombinator.com/item?id=50227330) · 44 points · 44 comments ## Similar posts on daily.dev - [Inflect-Micro-v2: complete voice in 9.36M parameters](https://daily.dev/posts/inflect-micro-v2-complete-voice-in-9-36m-parameters-yqa9jhasp) · Hacker News · 0 upvotes · 0 comments - [Voice Generation with TTS Model, Maya1](https://daily.dev/posts/voice-generation-with-tts-model-maya1-nqvpoksre) · DigitalOcean Community · 0 upvotes · 0 comments - [Open TTS Leaderboard: Scalable Evaluation for Multilingual Text-to-Speech and Voice Cloning](https://daily.dev/posts/open-tts-leaderboard-scalable-evaluation-for-multilingual-text-to-speech-and-voice-cloning-utmqtcomu) · Hugging Face · 0 upvotes · 0 comments - [Hugging Face Transformers in Action: Learning How To Leverage AI for NLP](https://daily.dev/posts/hugging-face-transformers-in-action-learning-how-to-leverage-ai-for-nlp-pwpvggwvt) · Towards Data Science · 0 upvotes · 0 comments --- Tags: [#deep-learning](https://daily.dev/tags/deep-learning), [#text-to-speech](https://daily.dev/tags/text-to-speech), [#huggingface](https://daily.dev/tags/huggingface), [#twilio](https://daily.dev/tags/twilio) [View this post on daily.dev](https://daily.dev/posts/show-hn-hurston-1-open-source-tts-that-speak-southern-american-and-afro-english-mccjqt4zw) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Show HN: Hurston-1 open source TTS that speak Southern American and Afro-English","url":"https://daily.dev/posts/show-hn-hurston-1-open-source-tts-that-speak-southern-american-and-afro-english-mccjqt4zw","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/show-hn-hurston-1-open-source-tts-that-speak-southern-american-and-afro-english-mccjqt4zw"},"datePublished":"2026-10-26T00:48:39.058Z","dateModified":"2026-10-26T00:49:18.767Z","description":"Hurston-1 is an open text-to-speech model fine-tuned from Maya1 to handle English accents most TTS systems get wrong, including Southern American, Nigerian,...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/aafca44af0ad607f2e86d3ccd1c46b62?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/aafca44af0ad607f2e86d3ccd1c46b62?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"Hacker News","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Hacker News","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/hn","url":"https://daily.dev/sources/hn"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/show-hn-hurston-1-open-source-tts-that-speak-southern-american-and-afro-english-mccjqt4zw","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"deep-learning,text-to-speech,huggingface,twilio","timeRequired":"PT11M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Hacker News","item":"https://daily.dev/sources/hn"},{"@type":"ListItem","position":3,"name":"Show HN: Hurston-1 open source TTS that speak Southern American and Afro-English"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/show-hn-hurston-1-open-source-tts-that-speak-southern-american-and-afro-english-mccjqt4zw#faq","mainEntity":[{"@type":"Question","name":"How do I generate 8kHz G.711 audio for Twilio from a text-to-speech model?","acceptedAnswer":{"@type":"Answer","text":"Hurston-1, a fine-tune of Maya1, has a built-in phone mode that filters output to the 300-3,400 Hz telephone band, resamples to 8 kHz, and encodes as G.711 mu-law (or A-law). Running the model with the --phone flag or phone=True produces this output, and ulaw_encode() returns raw mu-law bytes ready for a Twilio media stream. Developers wiring TTS into Twilio call flows can track tools like this on daily.dev."}},{"@type":"Question","name":"What open-source text-to-speech model handles African and Southern American English accents well?","acceptedAnswer":{"@type":"Answer","text":"Hurston-1, an Apache-2.0 fine-tune of Maya1, targets Southern American, Nigerian, West African, East African and Southern African English accents that most TTS systems handle poorly. In blind listening tests comparing it to base Maya1, Hurston-1 was preferred 78% of the time versus 6% for Maya1, which was never chosen for an African or Southern American voice. Developers evaluating accented speech models for inclusive voice products can follow releases like this on daily.dev."}},{"@type":"Question","name":"How much training data and compute does it take to fine-tune a TTS model like Maya1 for new accents?","acceptedAnswer":{"@type":"Answer","text":"Hurston-1 was produced with a LoRA fine-tune at rank 64, run for 2,000 steps, using about 190 hours of speech. Roughly 136 hours came from Destined AI's own accented English corpus, including about 10 hours of real phone calls with AI voice agents, supplemented by quality-filtered Mozilla Common Voice recordings from African-accented speakers. Teams planning their own LoRA fine-tunes can benchmark data and step counts against examples like this on daily.dev."}}]} ```