Destined AI open-sources Hurston-1 speech model
The Atlanta-based developer launched an open-source narrowband text-to-speech model fine-tuned on 136 hours of audio data.
Destined AI, an artificial intelligence company based in Atlanta, made its narrowband text-to-speech architecture Hurston-1 publicly accessible on Oct. 25 under an Apache-2.0 open-source license [1].
The software provides a dedicated narrowband speech synthesis tool created to generate synthetic speech [1]. According to an announcement shared by the project founder on Hacker News, the system was trained through a targeted fine-tuning process that incorporated 136 hours of recorded audio material [1]. The underlying code and model weights were made available under permissive Apache-2.0 distribution terms [1].
The developer focuses on building open-source narrowband text-to-speech systems that specialize in regional speech varieties, including Southern American accents and regional African English phrasing [1]. Under the open-source release parameters announced on Hacker News, external researchers and software developers can review, test, and implement the model weights within their own software stacks [1]. Destined AI has not publicly reported additional commercial terms, third-party distribution agreements, or future release timelines in connection with the Hurston-1 model announcement [1].
Questions and answers
What license does Hurston-1 use?
Hurston-1 is published under an open-source Apache-2.0 license, according to the founder's announcement on Hacker News.
How much audio was used to fine-tune Hurston-1?
The model was fine-tuned using 136 hours of audio data, according to the announcement.
Companies & investors in this story
Sources
- 1.Show HN: Hurston-1 open source TTS that speak Southern American and Afro-English — Hacker News, Oct 25, 2026
