Translate / app.py
Athspi's picture
Update app.py
21e6f34 verified
raw
history blame
6.91 kB
import os
import time
import tempfile
import uuid
import google.generativeai as genai
import requests
from flask import Flask, request, render_template, send_from_directory, url_for, flash
from moviepy.video.io.VideoFileClip import VideoFileClip
from moviepy.audio.io.AudioFileClip import AudioFileClip
from werkzeug.utils import secure_filename
from dotenv import load_dotenv
# --- 1. INITIALIZE FLASK APP AND LOAD SECRETS ---
load_dotenv()
app = Flask(__name__)
# Load secrets from the loaded environment variables
GEMINI_API_KEY = os.getenv("GEMINI_API_KEY")
TTS_API_URL = os.getenv("TTS_API_URL")
# Check if secrets were loaded correctly
if not GEMINI_API_KEY:
raise ValueError("SECURITY ERROR: GEMINI_API_KEY not found in .env file!")
if not TTS_API_URL:
raise ValueError("CONFIGURATION ERROR: TTS_API_URL not found in .env file!")
# Configure the Gemini API
genai.configure(api_key=GEMINI_API_KEY)
# Configure directories and app settings
UPLOAD_FOLDER = 'uploads'
DOWNLOAD_FOLDER = 'downloads'
os.makedirs(UPLOAD_FOLDER, exist_ok=True)
os.makedirs(DOWNLOAD_FOLDER, exist_ok=True)
app.config['UPLOAD_FOLDER'] = UPLOAD_FOLDER
app.config['DOWNLOAD_FOLDER'] = DOWNLOAD_FOLDER
app.config['MAX_CONTENT_LENGTH'] = 100 * 1024 * 1024 # 100 MB upload limit
app.secret_key = os.urandom(24) # Secure key for flash messages
# --- 2. VOICE CHOICES & GEMINI PROMPT ---
VOICE_CHOICES = {"Male (Charon)": "Charon", "Female (Zephyr)": "Zephyr"}
GEMINI_PROMPT = """
You are an expert AI scriptwriter. Your task is to watch the provided video and transcribe ALL spoken dialogue into a SINGLE, CONTINUOUS block of modern, colloquial Tamil.
**CRITICAL INSTRUCTIONS:**
1. **Single Script:** Combine all dialogue from all speakers into one continuous script.
2. **NO Timestamps or Speaker Labels:** Do NOT include any timestamps or speaker identifiers.
3. **Incorporate Performance:** Add English style prompts (e.g., `Say happily:`, `Whisper mysteriously:`) and performance tags (e.g., `[laugh]`, `[sigh]`) directly into the text for an expressive narration.
**EXAMPLE OUTPUT:**
Say happily: வணக்கம்! [laugh] எப்படி இருக்கீங்க? Whisper mysteriously: அந்த ரகசியம் எனக்கு மட்டும் தான் தெரியும்.
"""
# --- 3. CORE LOGIC HELPER FUNCTIONS ---
def generate_tamil_script(video_file_path):
"""Generates a single, continuous Tamil script from the video using Gemini."""
print("Uploading file to Gemini for transcription...")
video_file = genai.upload_file(video_file_path, mime_type="video/mp4")
print("Waiting for Gemini file processing...")
while video_file.state.name == "PROCESSING":
time.sleep(5)
video_file = genai.get_file(video_file.name)
if video_file.state.name != "ACTIVE":
raise Exception(f"Gemini file processing failed: {video_file.state.name}")
print("Generating narrator script...")
model = genai.GenerativeModel(model_name="models/gemini-2.5-flash")
response = model.generate_content([GEMINI_PROMPT, video_file])
genai.delete_file(video_file.name)
if hasattr(response, 'text') and response.text:
return " ".join(response.text.strip().splitlines())
raise Exception("No valid script was generated by Gemini.")
def generate_single_audio_track(dialogue_text, voice_name, is_cheerful, output_path):
"""Generates one audio track for the entire script via TTS API."""
print(f"Requesting audio from TTS API (Voice: {voice_name}, Cheerful: {is_cheerful})")
payload = {"text": dialogue_text, "voice_name": voice_name, "cheerful": is_cheerful}
response = requests.post(TTS_API_URL, json=payload, timeout=300)
if response.status_code == 200:
with open(output_path, "wb") as f:
f.write(response.content)
return True
raise Exception(f"Error from TTS API: {response.status_code} - {response.text}")
def replace_video_audio(video_path, new_audio_path, output_path):
"""Replaces the audio of a video using the modern, correct MoviePy method."""
print("Replacing video audio using MoviePy...")
video_clip, audio_clip = None, None
try:
video_clip = VideoFileClip(video_path)
audio_clip = AudioFileClip(new_audio_path)
video_clip.audio = audio_clip
video_clip.write_videofile(output_path, codec="libx264", audio_codec="aac", logger='bar')
finally:
if audio_clip: audio_clip.close()
if video_clip: video_clip.close()
# --- 4. FLASK WEB ROUTES ---
@app.route('/', methods=['GET'])
def index():
"""Renders the main upload page."""
return render_template('index.html')
@app.route('/process', methods=['POST'])
def process_video():
"""Handles video upload, processing, and renders the result."""
input_video_path, temp_audio_path = None, None
try:
if 'video' not in request.files or request.files['video'].filename == '':
flash("Please upload a video file.")
return render_template('index.html')
file = request.files['video']
filename = secure_filename(file.filename)
input_video_path = os.path.join(app.config['UPLOAD_FOLDER'], filename)
file.save(input_video_path)
voice_choice = request.form['voice_choice']
is_cheerful = 'cheerful' in request.form
voice_name = VOICE_CHOICES[voice_choice]
script = generate_tamil_script(input_video_path)
with tempfile.NamedTemporaryFile(suffix=".wav", delete=False) as temp_audio:
temp_audio_path = temp_audio.name
generate_single_audio_track(script, voice_name, is_cheerful, temp_audio_path)
final_video_name = f"dubbed_{filename}"
final_video_path = os.path.join(app.config['DOWNLOAD_FOLDER'], final_video_name)
replace_video_audio(input_video_path, temp_audio_path, final_video_path)
flash("Video processing complete!", "success")
return render_template('index.html',
result_video=url_for('serve_video', filename=final_video_name),
script=script)
except Exception as e:
print(f"An error occurred during processing: {e}")
flash(f"An unexpected error occurred: {e}. Please check the console and try again.", "error")
return render_template('index.html')
finally:
if input_video_path and os.path.exists(input_video_path):
os.remove(input_video_path)
if temp_audio_path and os.path.exists(temp_audio_path):
os.remove(temp_audio_path)
@app.route('/downloads/<filename>')
def serve_video(filename):
"""Serves the final dubbed video."""
return send_from_directory(app.config['DOWNLOAD_FOLDER'], filename)
# --- 5. APPLICATION ENTRY POINT ---
if __name__ == '__main__':
app.run(host="0.0.0.0", port=7860)