Introducing OhanTTS: Armenian Text-to-Speech
Blog post from Hugging Face
OhanTTS-v0.1 is a new Armenian-focused text-to-speech model that also supports English, Spanish, Arabic, Russian, Mandarin Chinese, Japanese, and French, trained on about 160 hours of recordings, including 61 hours of Armenian speech. It streams generated audio incrementally, with first audio arriving in about 331 milliseconds for a single client in the reported benchmark, and supports voice cloning from a 3–10-second recording and transcript, as well as voice creation from a written description. On held-out Armenian tests, it recorded a 14.70% word error rate, 6.15% character error rate, and 0.693 speaker-similarity score; the benchmark also found that generation remained faster than real time with up to 16 concurrent clients on the tested hardware. Released under the CC BY-NC 4.0 license for non-commercial use, the model currently limits requests to roughly 20 seconds and may struggle with Armenian punctuation, unfamiliar words, mixed-language speech, or poor reference recordings. Its developers plan to add streaming input in chunks and improve pronunciation and longer-form generation, and advise using cloned voices only with the speaker’s permission.
Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.