app.py

import os
import gradio as gr
import whisper
from moviepy.editor import VideoFileClip, AudioFileClip
from gtts import gTTS
from google import genai

# -----------------------
# Gemini Client Setup
# -----------------------
client = genai.Client(api_key=os.getenv("GEMINI_API_KEY"))

# -----------------------
# Load Whisper Model
# -----------------------
whisper_model = whisper.load_model("base")

def process_video(video_file):

    # Load video
    video = VideoFileClip(video_file)

    # Remove original audio (Safe Mode)
    muted_video = video.without_audio()
    muted_path = "muted_video.mp4"
    muted_video.write_videofile(muted_path, codec="libx264")

    # Extract audio
    audio_path = "temp_audio.wav"
    video.audio.write_audiofile(audio_path)

    # Transcribe
    result = whisper_model.transcribe(audio_path)
    transcript = result["text"]

    # Generate recap
    prompt = f"""
    Rewrite this transcript into a short Myanmar movie recap narration.
    Make it commentary style and transformative.
    Keep it under 3 minutes reading time.

    Transcript:
    {transcript}
    """

    response = client.models.generate_content(
        model="gemini-1.5-flash",
        contents=prompt,
    )

    recap_text = response.text

    # Convert to Myanmar voice
    tts = gTTS(recap_text, lang="my")
    voice_path = "recap_voice.mp3"
    tts.save(voice_path)

    # Merge new voice
    new_audio = AudioFileClip(voice_path)
    final_video = muted_video.set_audio(new_audio)
    output_path = "final_recap_video.mp4"
    final_video.write_videofile(output_path, codec="libx264", audio_codec="aac")

    return recap_text, output_path


interface = gr.Interface(
    fn=process_video,
    inputs=gr.Video(label="Upload Video (Max 5 min)"),
    outputs=[
        gr.Textbox(label="Myanmar Recap Text"),
        gr.Video(label="Final Recap Video")
    ],
    title="Auto Movie Recap Tool (Safe Mode)",
)

interface.launch()