AIZYS single synthetic Chinese speech sample. Test date 2026-10-22. macOS Tingting voice; no actual person or customer recording. faster-whisper 1.2.1; ctranslate2 4.8.3; numpy 1.26.4. CPU int8, threads=4, beam_size=5, temperature=0, language=zh, condition_on_previous_text=False. Uncorrected output is in asr.json. This single sample is not a model ranking.
