App.py

import streamlit as st
import openai
import torch
from transformers import pipeline
from deepface import DeepFace
import cv2
import numpy as np
import random
import os
from moviepy.editor import *
from faster_whisper import WhisperModel
from insightface.app import FaceAnalysis
import shutil
import tempfile

# Set Streamlit Page
st.set_page_config(page_title="DeepFlick: AI Face Swap & Voice Clone", layout="wide")

st.title("🎭 DeepFlick: AI-Powered Face Swap & Voice Clone")

# Load Face Swap Model
face_app = FaceAnalysis(name="buffalo_l")
face_app.prepare(ctx_id=0, det_size=(640, 640))

# Text-to-Speech Model
tts_pipeline = pipeline("text-to-speech", model="coqui-ai/TTS")

# Whisper Model for Auto-Captions
whisper_model = WhisperModel("base", device="cuda" if torch.cuda.is_available() else "cpu")

# Function to Swap Faces
def swap_faces(source_img_path, target_img_path, output_path):
    source_img = cv2.imread(source_img_path)
    target_img = cv2.imread(target_img_path)

    # Detect faces
    source_faces = face_app.get(source_img)
    target_faces = face_app.get(target_img)

    if len(source_faces) == 0 or len(target_faces) == 0:
        return None

    # Swap faces
    swapped_img = DeepFace.swap(source_img, target_img, model_name="insightface")

    # Save output
    cv2.imwrite(output_path, swapped_img)
    return output_path

# Function to Generate Voiceover
def generate_voice(script_text):
    audio_path = f"output_{random.randint(1000,9999)}.wav"
    tts_audio = tts_pipeline(script_text)
    
    with open(audio_path, "wb") as f:
        f.write(tts_audio["audio"])
    
    return audio_path

# Function to Generate Captions
def generate_captions(audio_path):
    segments, _ = whisper_model.transcribe(audio_path)
    captions = "\n".join([segment.text for segment in segments])
    return captions

# User Inputs
st.sidebar.subheader("🎭 Upload Images")
source_img = st.sidebar.file_uploader("Upload Your Face (Source)", type=["jpg", "png"])
target_img = st.sidebar.file_uploader("Upload Target Face", type=["jpg", "png"])

st.sidebar.subheader("🎙️ AI Voice Options")
script_text = st.sidebar.text_area("Enter Video Script:", "Type something interesting here...")
use_ai_voice = st.sidebar.checkbox("Use AI Voice Clone", value=True)

# Process Button
if st.sidebar.button("🎬 Generate AI Video"):
    if source_img and target_img and script_text:
        with st.spinner("Processing AI Face Swap & Voice Clone... 🚀"):
            # Save temp images
            temp_source = tempfile.NamedTemporaryFile(delete=False, suffix=".jpg")
            temp_target = tempfile.NamedTemporaryFile(delete=False, suffix=".jpg")
            temp_source.write(source_img.read())
            temp_target.write(target_img.read())

            # Perform Face Swap
            output_img_path = "swapped_face.jpg"
            swapped_face = swap_faces(temp_source.name, temp_target.name, output_img_path)

            if swapped_face:
                st.image(swapped_face, caption="AI-Swapped Face", use_column_width=True)

                # Generate AI Voice
                if use_ai_voice:
                    audio_path = generate_voice(script_text)
                    st.audio(audio_path, format="audio/wav")

                    # Generate Captions
                    captions = generate_captions(audio_path)
                    st.write("📝 **Auto-Generated Captions:**", captions)

                # Convert Swapped Image to Video
                video_clip = ImageSequenceClip([swapped_face] * 30, fps=10)
                video_audio = AudioFileClip(audio_path) if use_ai_voice else None
                final_video_path = f"deepflick_output_{random.randint(1000,9999)}.mp4"

                if video_audio:
                    video_clip = video_clip.set_audio(video_audio)

                video_clip.write_videofile(final_video_path, codec="libx264")

                st.video(final_video_path)

            else:
                st.error("Face swap failed! Try using clearer images.")

            # Cleanup temp files
            temp_source.close()
            temp_target.close()
            os.remove(temp_source.name)
            os.remove(temp_target.name)

    else:
        st.warning("⚠️ Please upload both images and enter a script.")

# Search Functionality
st.markdown("---")
st.subheader("🔍 Search AI-Generated Videos")

search_query = st.text_input("Search for a video by script content:")
if st.button("🔍 Search"):
    # Simulated Search (Replace with a database search)
    dummy_videos = {
        "Funny AI Skit": "deepflick_output_1234.mp4",
        "AI Meme Video": "deepflick_output_5678.mp4"
    }

    found_videos = {title: path for title, path in dummy_videos.items() if search_query.lower() in title.lower()}
    
    if found_videos:
        for title, path in found_videos.items():
            st.markdown(f"### 🎥 {title}")
            st.video(path)
    else:
        st.write("❌ No videos found matching your search.")