Speech Recognition WhisperX (Multi Speaker) pairs the local WhisperX engine with pyannote speaker diarization to handle recordings with multiple participants. Upload meetings, interviews, panel discussions, or customer-support calls and the system automatically labels each segment with SPEAKER_01 / SPEAKER_02 / SPEAKER_03, so reviewers can immediately see who said what. Outputs include a plain transcript, SRT subtitles, and a diarized transcript. The full audio is processed in one request, which keeps SPEAKER labels consistent across the whole recording (no sudden re-numbering halfway through). The pipeline runs entirely on your on-prem GPU, making it suitable for internal meeting minutes, legal records, and medical consultations. Mixed Chinese / English and common Asian languages are supported. Outputs feed directly into downstream summarization, translation, or meeting-note workflows. For single-speaker recordings, use "Speech Recognition WhisperX (Single Speaker)" instead for faster processing.