app.py

import gradio as gr
from tqdm.asyncio import tqdm_asyncio

from pipeline import (
    extract_audio_from_video,
    transcribe_and_preprocess_audio,
    translation_main,
    tts_main,
    create_combined_output,
)
from pipeline import translation_hdr, translation_url, LANG


async def process_video_translation(
    input_video, speaker, progress=gr.Progress(track_tqdm=True)
):
    if input_video is None:
        gr.Info("Please upload a video file", duration=2)
        return

    total_stages = 6

    output_video = f"{input_video.split('.')[0]}_translated.mp4"
    with tqdm_asyncio(total=total_stages, desc="Processing video translation") as pbar:

        # stage 1: extract audio from video
        progress(0.1, desc="Extracting audio from video")
        output_audio_path = extract_audio_from_video(input_video)
        pbar.update(1)

        # transcribe audio
        progress(0.2, desc="Transcribing audio")
        sentences = transcribe_and_preprocess_audio(output_audio_path)
        pbar.update(1)

        # translate to twi
        progress(0.4, desc="Translating to Twi")
        khaya_translations = await translation_main(
            sentences, translation_url, translation_hdr, LANG
        )
        pbar.update(1)

        # convert to speech
        progress(0.7, desc="Converting to speech")
        output_audio = await tts_main(khaya_translations, speaker)
        # print(tts_output_files)
        pbar.update(1)

        progress(1.0, desc="Combining audio and video")
        create_combined_output(input_video, output_audio, output_video)
        pbar.update(1)

        print("Video translation completed")
        gr.Info(f"Video translation completed", duration=2)

        return output_video


app_theme = gr.themes.Ocean(
    text_size="lg",
    spacing_size="lg",
)
with gr.Blocks(
    theme=app_theme,
    title="Video Dubbing Interface",
) as demo:
    with gr.Row(variant="default"):
        with gr.Column(
            scale=1,
            min_width=0,
        ):
            gr.Image(
                "logo_2.jpeg",
                show_label=False,
                height=200,
                show_download_button=False,
                show_fullscreen_button=False,
                container=False,
                show_share_button=False,
            )
        with gr.Column(
            scale=6,
            variant="default",
        ):
            gr.HTML(
                """
                <h1 style="font-size: 4em; font-weight: bold; margin-top: 0.5em; margin-left:3em">
                    Video Dubbing Interface
                </h1>

                """,
            )
        with gr.Column(
            scale=1,
            min_width=0,
        ):
            gr.Image(
                "NLPGhana_logo_2.png",
                show_label=False,
                height=200,
                show_download_button=False,
                show_fullscreen_button=False,
                container=False,
                show_share_button=False,
            )
    gr.HTML("<hr style='margin-top: 0.5em;'>")

    gr.HTML("<div style='height: 20px;'></div>")

    # main interface components
    with gr.Row():
        with gr.Column():
            input_video = gr.Video(label="Input Video", sources=["upload"], height=400)
            input_speaker = gr.Radio(
                label="Select Speaker",
                choices=["male", "female"],
                value="female",
                min_width=50,
                container=True,
                show_label=True,
            )
            submit = gr.Button("Process Video", scale=1)
        output_video = gr.Video(label="Processed Video", height=400)
        submit.click(
            process_video_translation,
            inputs=[input_video, input_speaker],
            outputs=output_video,
        )

    gr.HTML("<div style='height: 10px;'></div>")


# Launch the interface
demo.launch(debug=True)