Its Hugging Face model card specifies 48 kHz output and says it runs faster than real time on a CPU. The code and weights are licensed under Apache-2.0, and the package can be installed with pip install antalia-mini.
The model card describes an included text normalizer for numbers, dates, times, amounts, abbreviations, URLs, and English or brand terms. It also documents batched and streaming synthesis, configurable output rates, and a browser demo that runs locally with ONNX Runtime Web. The voice is a single synthetic male voice; the card says the training data contains no recordings of real speakers.
The published evaluation compares Antalia-2 Mini with EMA Lightning and other systems on Turkish intelligibility and naturalness measures. The authors report similar WER results for Antalia-2 Mini and EMA Lightning in their harness, while the naturalness scores differ by metric and audio band. These are model-card results, not an independent evaluation.