Show HN: Hurston-1 open source TTS that speak Southern American and Afro-English
AI model · https://huggingface.co/destined/hurston-1
First seen by Oriel on .
AI summary
Hurston-1 is an open Maya1 fine-tune for accented English speech, with phone-ready 8 kHz G.711 output and voice steering through plain-language descriptions.
- It targets Southern American and African English accents, generates a new voice for each call, and does not clone real people.
- The model was fine-tuned for 2,000 steps on about 190 hours of speech, including roughly 10 hours from real calls with AI voice agents.
- Phone mode filters speech to the telephone band, resamples it to 8 kHz, and encodes G.711 mu-law or A-law audio for telephony systems.
- On 510 harder clips, mean word error rate fell from 0.145 in single takes to 0.062 with --check; names and numbers remained the weakest category.
- Intended uses include accent-coverage testing, simulated callers, training-data augmentation, and voice prompts; the page advises human review before important use.
#text-to-speech #accented english #telephony #speech synthesis
Generated by gpt-5.6-sol. Check the original for details.
Metric trends
Last 30 days · Up to 120 recorded samples per source and metric, with reposts kept separate · First and latest retained
Points
Score reported by the source.
- Hacker News: 65, latest observation ; +59 over observed period since
Comments
Comment count reported by the source.
- Hacker News: 25, latest observation ; +24 over observed period since
Aliases
| Source | External ID | URL |
|---|---|---|
| Hacker News | 50227330 | https://huggingface.co/destined/hurston-1 Read discussion |