jdt-pxl/meta-taiwanese-hokkien-english-s2ut

Taiwanese Hokkien → English S2ST

Demo on Meta's Taiwanese Hokkien → English speech-to-speech translation (S2UT) model.

Model: facebook/xm_transformer_s2ut_hk-en

Usage

Upload a .wav file (or record directly) containing Philippine Hokkien speech. The model outputs synthesised English audio and its ASR transcript.

⚠️ This Space runs on CPU. Inference takes ~60–120 s per clip. Keep inputs under 10 s for best results.