What changed

Hugging Face announced OhanTTS, an Armenian-focused text-to-speech (TTS) model that generates natural speech in Armenian and seven other languages: English, Spanish, Arabic, Russian, Mandarin Chinese, Japanese, and French. The first release, OhanTTS-v0.1, is built on a multilingual speech model and was trained on approximately 160 hours of reference recordings across these eight languages, including 61 hours of Armenian speech. OhanTTS supports reference voice cloning, allowing users to generate speech in a specific voice by providing a 3–10 second audio sample and its transcript.

What this means in practice

OhanTTS streams audio incrementally, enabling users to hear the first words in about a third of a second. This reduces latency compared to traditional TTS models that generate the entire utterance before returning audio. The model is served through SGLang-Omni with an OpenAI-compatible /v1/audio/speech endpoint. It is available under the Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0) license, making it freely available for research, educational, and non-commercial use. The model can also create a voice from a short description without a reference recording, and it supports generating speech in English.

Limitations and availability

OhanTTS is an early release with some known limitations. Long utterances are capped at approximately 20 seconds of audio, and longer inputs should be split. Armenian punctuation marks may not always be reflected correctly in pauses, stress, and intonation. Unfamiliar names and foreign words may be mispronounced, and code-switching between languages within an utterance can occasionally cause errors. Reference quality depends on a clean recording and an exact transcript, and users should obtain permission before using reference voices. The model is available for non-commercial use and can be tried through the provided endpoint.

Source facts for Hugging Face Launches OhanTTS: Armenian-Focused Text-to-Speech Model: Alongside Armenian, it generates speech in English, Spanish, Arabic, Russian, Mandarin Chinese, Japanese, and French. OhanTTS supports reference voice cloning: provide a 3–10 second audio sample and its transcript, and the model generates speech in that voice. His voice reached beyond the mountains.
Original AINewsTrend explainer based on the cited source. This graphic summarizes source statements; it is not an independent benchmark.

Sources and further reading

  1. huggingface: Introducing OhanTTS: Armenian Text-to-Speech