app.py

import os
import gradio as gr
import whisperx

HF_TOKEN = os.getenv("HF_TOKEN")  # MUST be set in HF Spaces secrets

ASR_MODEL = "palli23/whisper-small-sam_spjall-ct2"
DIARIZATION_MODEL = "pyannote/speaker-diarization-3.1"
ALIGN_MODEL = "WAV2VEC2_ASR_LARGE_LV60K_960H"

def load_models():
    print("Loading WhisperX ASR...")
    asr = whisperx.load_model(
        model_name=ASR_MODEL,
        device="cuda" if whisperx.is_cuda_available() else "cpu",
        compute_type="int8",  # Safe for Spaces
        hf_token=HF_TOKEN
    )

    print("Loading alignment model...")
    align_model, metadata = whisperx.load_align_model(
        language_code="is", 
        model_name=ALIGN_MODEL,
        hf_token=HF_TOKEN
    )

    print("Loading diarization model...")
    diar = whisperx.DiarizationPipeline(
        DIARIZATION_MODEL,
        hf_token=HF_TOKEN,
        use_auth_token=True
    )

    return asr, align_model, metadata, diar


asr_model, align_model, align_metadata, diar_pipeline = load_models()


def transcribe(audio):
    if audio is None:
        return "No audio provided."

    print("Running ASR...")
    result = asr_model.transcribe(audio)

    print("Running alignment...")
    aligned = whisperx.align(
        result["segments"],
        align_model,
        align_metadata,
        audio,
        "is"
    )

    print("Running diarization...")
    diarization = diar_pipeline(audio)

    print("Assigning speaker labels...")
    final_result = whisperx.assign_word_speakers(
        diarization,
        aligned
    )

    text_out = ""
    for seg in final_result["segments"]:
        speaker = seg.get("speaker", "Unknown")
        text_out += f"[{speaker}] {seg['text']}\n"

    return text_out


ui = gr.Interface(
    fn=transcribe,
    inputs=gr.Audio(type="filepath"),
    outputs=gr.Textbox(label="Transcription + Speakers", lines=20),
    title="WhisperX Icelandic CT2 + Diarization",
    description="Uses your private CT2 Whisper Small model + alignment + pyannote diarization."
)

if __name__ == "__main__":
    ui.launch()