stt▌

98%+ accurate transcription with speaker diarization, audio event tagging, and word-level forced alignment. \n \n Supports Scribe v1 and v2 models with auto-detection across 90+ languages \n Capabilities include speaker identification, audio event tagging (laughter, applause, music), and precise word-level timestamps via forced alignment \n Forced alignment enables subtitle generation, lip-sync timing, and karaoke applications by aligning known text to audio \n Requires inference.sh CLI ( infsh