urdu_tts_video.py

import gradio as gr
import os
import cv2
import ffmpeg
import numpy as np
from gtts import gTTS
from diffusers import StableDiffusionPipeline
import torch

# Ensure required folders exist
os.makedirs("generated_images", exist_ok=True)
os.makedirs("output", exist_ok=True)

# Load Stable Diffusion for image generation
model_id = "runwayml/stable-diffusion-v1-5"
pipe = StableDiffusionPipeline.from_pretrained(model_id, torch_dtype=torch.float32)
pipe.safety_checker = None  # Disable safety checker

# Global variable to store generated TTS audio path
global_audio_path = None


### 🗣️ TEXT-TO-SPEECH FUNCTION ###
def text_to_speech(script_file):
    if script_file is None:
        return None, "⚠️ Please upload an Urdu script file!"
    
    with open(script_file.name, "r", encoding="utf-8") as f:
        urdu_text = f.read().strip()

    audio_path = "output/urdu_audio.mp3"
    tts = gTTS(text=urdu_text, lang="ur")
    tts.save(audio_path)

    global global_audio_path
    global_audio_path = audio_path

    return audio_path, "✅ Audio generated successfully!"


### 🏞️ IMAGE GENERATION FUNCTION ###
def generate_images(script_file, num_images):
    if script_file is None:
        return None, "⚠️ Please upload a script file!"
    
    num_images = int(num_images)

    with open(script_file.name, "r", encoding="utf-8") as f:
        text_lines = f.read().split("\n\n")  # Splitting scenes by double newlines
    
    image_paths = []
    for i, scene in enumerate(text_lines[:num_images]):
        prompt = f"Scene {i+1}: {scene.strip()}"
        image = pipe(prompt).images[0]
        image_path = f"generated_images/image_{i+1}.png"
        image.save(image_path)
        image_paths.append(image_path)
    
    return image_paths, "✅ Images generated successfully!"


### 🎥 VIDEO CREATION FUNCTION ###
def images_to_video(image_paths, fps=1):
    if not image_paths:
        return None
    
    frame = cv2.imread(image_paths[0])
    height, width, layers = frame.shape

    video_path = "output/generated_video.mp4"
    fourcc = cv2.VideoWriter_fourcc(*"mp4v")
    video = cv2.VideoWriter(video_path, fourcc, fps, (width, height))

    for image in image_paths:
        frame = cv2.imread(image)
        video.write(frame)
    
    video.release()
    return video_path


### 🔊 AUDIO-VIDEO MERGE FUNCTION ###
def merge_audio_video(video_path):
    if global_audio_path is None:
        return None, "⚠️ No audio found! Please generate Urdu TTS first."
    
    final_video_path = "output/final_video.mp4"
    
    video = ffmpeg.input(video_path)
    audio = ffmpeg.input(global_audio_path)
    
    ffmpeg.output(video, audio, final_video_path, vcodec="libx264", acodec="aac").run(overwrite_output=True)
    
    return final_video_path, "✅ Video with Urdu voice-over generated successfully!"


### 🎬 FINAL VIDEO GENERATION PIPELINE ###
def generate_final_video(script_file, num_images):
    if script_file is None:
        return None, "⚠️ Please upload a script file for image generation!"
    
    image_paths, img_msg = generate_images(script_file, num_images)
    if not image_paths:
        return None, img_msg

    video_path = images_to_video(image_paths, fps=1)
    final_video_path, vid_msg = merge_audio_video(video_path)
    
    return final_video_path, vid_msg


### 🚀 GRADIO UI ###
with gr.Blocks() as demo:
    gr.Markdown("## 🎤 Urdu Text-to-Speech & AI Video Generator")
    
    # TTS Section
    with gr.Tab("🗣️ Urdu Text-to-Speech"):
        script_file_tts = gr.File(label="📂 Upload Urdu Script for Audio", type="filepath")
        generate_audio_btn = gr.Button("🎙️ Generate Audio", variant="primary")
        audio_output = gr.Audio(label="🔊 Urdu Speech Output", interactive=False)
        audio_status = gr.Textbox(label="ℹ️ Status", interactive=False)
    
    generate_audio_btn.click(text_to_speech, inputs=[script_file_tts], outputs=[audio_output, audio_status])
    
    # Video Generation Section
    with gr.Tab("🎥 AI Video Generator"):
        script_file_video = gr.File(label="📂 Upload Urdu Script for Images", type="filepath")
        num_images = gr.Number(label="📸 Number of Scenes", value=3, minimum=1, maximum=10, step=1)
        generate_video_btn = gr.Button("🎬 Generate Video", variant="primary")
        video_output = gr.Video(label="🎞️ Generated Video")
        video_status = gr.Textbox(label="ℹ️ Status", interactive=False)
    
    generate_video_btn.click(generate_final_video, inputs=[script_file_video, num_images], outputs=[video_output, video_status])

demo.launch()