🎙️ MeetPilot AI: Speech Recognition & Speaker Diarization

Transcribe audio and identify distinct speakers in a single pass using faster-whisper (large-v3) + pyannote.audio (community-1) with WhisperX alignment on Hugging Face ZeroGPU.

API Usage from MeetPilot Backend

This Space exposes a named API endpoint transcribe. You can call it programmatically via gradio_client or HTTP POST:

from gradio_client import Client, handle_file

client = Client("Subham05x/meetpilot-whisper-diarization-space", hf_token=os.environ["HF_TOKEN"])
result = client.predict(
    audio_file=handle_file("meeting.wav"),
    min_speakers=None,
    max_speakers=None,
    language=None,
    api_name="/transcribe"
)
print(result) # {'segments': [...], 'language': 'en', 'duration': 12.5}