app.py

import gradio as gr
from pipeline.utils import extract_audio
from pipeline.asr_diarization import transcribe_and_align
from pipeline.translation import translate_text
from pipeline.voice_synth import synthesize_speech
from pipeline.mixing import mix_speech_with_background
from pipeline.mux import mux_audio_video

def full_pipeline(video_path: str, target_language="hi"):
    # video_path is already a string path; don't use .name
    audio_path = extract_audio(video_path, out_audio="audio.wav")
    _, aligned = transcribe_and_align(audio_path, device="cpu")

    speech_segments = []
    for seg in aligned["segments"]:
        src_text = seg["text"]
        translated = translate_text(src_text)  # currently hardcoded en→hi
        synth_path = synthesize_speech(translated, speaker_id="speaker1")
        start_ms = int(seg["start"] * 1000)
        speech_segments.append({"file": synth_path, "start_ms": start_ms})

    final_audio = mix_speech_with_background(audio_path, speech_segments)
    output_video = mux_audio_video(video_path, final_audio)
    return output_video

with gr.Blocks() as demo:
    gr.Markdown("## AI Dubbing Prototype (translation + placeholder voice)")
    video_in = gr.Video(label="Upload short video clip")
    lang = gr.Textbox(value="hi", label="Target language (currently fixed)")
    out_video = gr.Video(label="Dubbed output")
    btn = gr.Button("Run Dubbing Pipeline")
    btn.click(fn=full_pipeline, inputs=[video_in, lang], outputs=[out_video])
demo.launch()