# Import necessary libraries from kokoro import KPipeline import soundfile as sf import torch import soundfile as sf import os from moviepy.editor import VideoFileClip, AudioFileClip, ImageClip from PIL import Image import tempfile import random import cv2 import math import os, requests, io, time, re, random from moviepy.editor import ( VideoFileClip, concatenate_videoclips, AudioFileClip, ImageClip, CompositeVideoClip, TextClip, CompositeAudioClip ) import gradio as gr import shutil import os import moviepy.video.fx.all as vfx import moviepy.config as mpy_config from pydub import AudioSegment from pydub.generators import Sine from PIL import Image, ImageDraw, ImageFont import numpy as np from bs4 import BeautifulSoup import base64 from urllib.parse import quote import pysrt from gtts import gTTS import gradio as gr # Import Gradio # Initialize Kokoro TTS pipeline (using American English) pipeline = KPipeline(lang_code='a') # Use voice 'af_heart' for American English # Ensure ImageMagick binary is set mpy_config.change_settings({"IMAGEMAGICK_BINARY": "/usr/bin/convert"}) # ---------------- Global Configuration ---------------- # PEXELS_API_KEY = 'BhJqbcdm9Vi90KqzXKAhnEHGsuFNv4irXuOjWtT761U49lRzo03qBGna' OPENROUTER_API_KEY = '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' OPENROUTER_MODEL = "mistralai/mistral-small-3.1-24b-instruct:free" OUTPUT_VIDEO_FILENAME = "final_video.mp4" USER_AGENT = "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/91.0.4472.124 Safari/537.36" # Additional global variables needed for the Gradio interface selected_voice = 'af_heart' # Default voice voice_speed = 0.9 # Default voice speed font_size = 45 # Default font size video_clip_probability = 0.25 # Default probability for video clips bg_music_volume = 0.08 # Default background music volume fps = 30 # Default FPS preset = "veryfast" # Default preset TARGET_RESOLUTION = None CAPTION_COLOR = None TEMP_FOLDER = None # ---------------- Helper Functions ---------------- # # (Your existing helper functions remain unchanged: generate_script, parse_script, # search_pexels_videos, search_pexels_images, search_google_images, download_image, # download_video, generate_media, generate_tts, apply_kenburns_effect, # resize_to_fill, find_mp3_files, add_background_music, create_clip, # fix_imagemagick_policy) # Define these globally as they were in your original code but will be set per run TARGET_RESOLUTION = None CAPTION_COLOR = None TEMP_FOLDER = None def generate_script(user_input): """Generate documentary script with proper OpenRouter handling.""" headers = { 'Authorization': f'Bearer {OPENROUTER_API_KEY}', 'HTTP-Referer': 'https://your-domain.com', 'X-Title': 'AI Documentary Maker' } prompt = f"""Short Documentary Script GeneratorInstructions: If I say "use this," just output the script exactly as I gave it. If I only give topics, generate a script based on them. If I provide a full script, rewrite it without any changes. Make everything short simple and humarous funny and act as serious but humarous. And don't say anything off topic. Also alway say a funny statement to subscribe based on the video topic at the end. Use normal conversational text like a normal person talking and avoid AI phase make the statements humanize and normal conversational And always talk negative views don't really wants any positive statement and the full script will be one video about one topic so keep in mind Formatting Rules: Title in Square Brackets: Each section starts with a one-word title inside [ ] (max two words if necessary). This title will be used as a search term for Pexels footage. Casual & Funny Narration: Each section has 5-10 words of narration. Keep it natural, funny, and unpredictable (not robotic, poetic, or rhythmic). No Special Formatting: No bold, italics, or special characters. You are a assistant AI your task is to create script. You aren't a chatbot. So, don't write extra text Generalized Search Terms: If a term is too specific, make it more general for Pexels search. Scene-Specific Writing: Each section describes only what should be shown in the video. Output Only the Script, and also make it funny and humarous and helirous and also add to subscribe with a funny statement like subscribe now or ..... No extra text, just the script. Example Output: [North Korea] Top 5 unknown facts about North Korea. [Invisibility] North Korea’s internet speed is so fast… it doesn’t exist. [Leadership] Kim Jong-un once won an election with 100% votes… against himself. [Magic] North Korea discovered time travel. That’s why their news is always from the past. [Warning] Subscribe now, or Kim Jong-un will send you a free one-way ticket… to North Korea. [Freedom] North Korean citizens can do anything… as long as it's government-approved. Now here is the Topic/scrip: {user_input} """ data = { 'model': OPENROUTER_MODEL, 'messages': [{'role': 'user', 'content': prompt}], 'temperature': 0.4, 'max_tokens': 5000 } try: response = requests.post( 'https://openrouter.ai/api/v1/chat/completions', headers=headers, json=data, timeout=30 ) if response.status_code == 200: response_data = response.json() if 'choices' in response_data and len(response_data['choices']) > 0: return response_data['choices'][0]['message']['content'] else: print("Unexpected response format:", response_data) return None else: print(f"API Error {response.status_code}: {response.text}") return None except Exception as e: print(f"Request failed: {str(e)}") return None def parse_script(script_text): """ Parse the generated script into a list of elements. For each section, create two elements: - A 'media' element using the section title as the visual prompt. - A 'tts' element with the narration text, voice info, and computed duration. """ sections = {} current_title = None current_text = "" try: for line in script_text.splitlines(): line = line.strip() if line.startswith("[") and "]" in line: bracket_start = line.find("[") bracket_end = line.find("]", bracket_start) if bracket_start != -1 and bracket_end != -1: if current_title is not None: sections[current_title] = current_text.strip() current_title = line[bracket_start+1:bracket_end] current_text = line[bracket_end+1:].strip() elif current_title: current_text += line + " " if current_title: sections[current_title] = current_text.strip() elements = [] for title, narration in sections.items(): if not title or not narration: continue media_element = {"type": "media", "prompt": title, "effects": "fade-in"} words = narration.split() duration = max(3, len(words) * 0.5) tts_element = {"type": "tts", "text": narration, "voice": "en", "duration": duration} elements.append(media_element) elements.append(tts_element) return elements except Exception as e: print(f"Error parsing script: {e}") return [] def search_pexels_videos(query, pexels_api_key): """Search for a video on Pexels by query and return a random HD video.""" headers = {'Authorization': pexels_api_key} base_url = "https://api.pexels.com/videos/search" num_pages = 3 videos_per_page = 15 max_retries = 3 retry_delay = 1 search_query = query all_videos = [] for page in range(1, num_pages + 1): for attempt in range(max_retries): try: params = {"query": search_query, "per_page": videos_per_page, "page": page} response = requests.get(base_url, headers=headers, params=params, timeout=10) if response.status_code == 200: data = response.json() videos = data.get("videos", []) if not videos: print(f"No videos found on page {page}.") break for video in videos: video_files = video.get("video_files", []) for file in video_files: if file.get("quality") == "hd": all_videos.append(file.get("link")) break break elif response.status_code == 429: print(f"Rate limit hit (attempt {attempt+1}/{max_retries}). Retrying in {retry_delay} seconds...") time.sleep(retry_delay) retry_delay *= 2 else: print(f"Error fetching videos: {response.status_code} {response.text}") if attempt < max_retries - 1: print(f"Retrying in {retry_delay} seconds...") time.sleep(retry_delay) retry_delay *= 2 else: break except requests.exceptions.RequestException as e: print(f"Request exception: {e}") if attempt < max_retries - 1: print(f"Retrying in {retry_delay} seconds...") time.sleep(retry_delay) retry_delay *= 2 else: break if all_videos: random_video = random.choice(all_videos) print(f"Selected random video from {len(all_videos)} HD videos") return random_video else: print("No suitable videos found after searching all pages.") return None def search_pexels_images(query, pexels_api_key): """Search for an image on Pexels by query.""" headers = {'Authorization': pexels_api_key} url = "https://api.pexels.com/v1/search" params = {"query": query, "per_page": 5, "orientation": "landscape"} max_retries = 3 retry_delay = 1 for attempt in range(max_retries): try: response = requests.get(url, headers=headers, params=params, timeout=10) if response.status_code == 200: data = response.json() photos = data.get("photos", []) if photos: photo = random.choice(photos[:min(5, len(photos))]) img_url = photo.get("src", {}).get("original") return img_url else: print(f"No images found for query: {query}") return None elif response.status_code == 429: print(f"Rate limit hit (attempt {attempt+1}/{max_retries}). Retrying in {retry_delay} seconds...") time.sleep(retry_delay) retry_delay *= 2 else: print(f"Error fetching images: {response.status_code} {response.text}") if attempt < max_retries - 1: print(f"Retrying in {retry_delay} seconds...") time.sleep(retry_delay) retry_delay *= 2 except requests.exceptions.RequestException as e: print(f"Request exception: {e}") if attempt < max_retries - 1: print(f"Retrying in {retry_delay} seconds...") time.sleep(retry_delay) retry_delay *= 2 print(f"No Pexels images found for query: {query} after all attempts") return None def search_google_images(query): """Search for images on Google Images (for news-related queries)""" try: search_url = f"https://www.google.com/search?q={quote(query)}&tbm=isch" headers = {"User-Agent": USER_AGENT} response = requests.get(search_url, headers=headers, timeout=10) soup = BeautifulSoup(response.text, "html.parser") img_tags = soup.find_all("img") image_urls = [] for img in img_tags: src = img.get("src", "") if src.startswith("http") and "gstatic" not in src: image_urls.append(src) if image_urls: return random.choice(image_urls[:5]) if len(image_urls) >= 5 else image_urls[0] else: print(f"No Google Images found for query: {query}") return None except Exception as e: print(f"Error in Google Images search: {e}") return None def download_image(image_url, filename): """Download an image from a URL to a local file with enhanced error handling.""" try: headers = {"User-Agent": USER_AGENT} print(f"Downloading image from: {image_url} to {filename}") response = requests.get(image_url, headers=headers, stream=True, timeout=15) response.raise_for_status() with open(filename, 'wb') as f: for chunk in response.iter_content(chunk_size=8192): f.write(chunk) print(f"Image downloaded successfully to: {filename}") try: img = Image.open(filename) img.verify() img = Image.open(filename) if img.mode != 'RGB': img = img.convert('RGB') img.save(filename) print(f"Image validated and processed: {filename}") return filename except Exception as e_validate: print(f"Downloaded file is not a valid image: {e_validate}") if os.path.exists(filename): os.remove(filename) return None except requests.exceptions.RequestException as e_download: print(f"Image download error: {e_download}") if os.path.exists(filename): os.remove(filename) return None except Exception as e_general: print(f"General error during image processing: {e_general}") if os.path.exists(filename): os.remove(filename) return None def download_video(video_url, filename): """Download a video from a URL to a local file.""" try: response = requests.get(video_url, stream=True, timeout=30) response.raise_for_status() with open(filename, 'wb') as f: for chunk in response.iter_content(chunk_size=8192): f.write(chunk) print(f"Video downloaded successfully to: {filename}") return filename except Exception as e: print(f"Video download error: {e}") if os.path.exists(filename): os.remove(filename) return None def generate_media(prompt, user_image=None, current_index=0, total_segments=1): """ Generate a visual asset by first searching for a video or using a specific search strategy. For news-related queries, use Google Images. Returns a dict: {'path': , 'asset_type': 'video' or 'image'}. """ safe_prompt = re.sub(r'[^\w\s-]', '', prompt).strip().replace(' ', '_') if "news" in prompt.lower(): print(f"News-related query detected: {prompt}. Using Google Images...") image_file = os.path.join(TEMP_FOLDER, f"{safe_prompt}_news.jpg") image_url = search_google_images(prompt) if image_url: downloaded_image = download_image(image_url, image_file) if downloaded_image: print(f"News image saved to {downloaded_image}") return {"path": downloaded_image, "asset_type": "image"} else: print(f"Google Images search failed for prompt: {prompt}") if random.random() < video_clip_probability: video_file = os.path.join(TEMP_FOLDER, f"{safe_prompt}_video.mp4") video_url = search_pexels_videos(prompt, PEXELS_API_KEY) if video_url: downloaded_video = download_video(video_url, video_file) if downloaded_video: print(f"Video asset saved to {downloaded_video}") return {"path": downloaded_video, "asset_type": "video"} else: print(f"Pexels video search failed for prompt: {prompt}") image_file = os.path.join(TEMP_FOLDER, f"{safe_prompt}.jpg") image_url = search_pexels_images(prompt, PEXELS_API_KEY) if image_url: downloaded_image = download_image(image_url, image_file) if downloaded_image: print(f"Image asset saved to {downloaded_image}") return {"path": downloaded_image, "asset_type": "image"} else: print(f"Pexels image download failed for prompt: {prompt}") fallback_terms = ["nature", "people", "landscape", "technology", "business"] for term in fallback_terms: print(f"Trying fallback image search with term: {term}") fallback_file = os.path.join(TEMP_FOLDER, f"fallback_{term}.jpg") fallback_url = search_pexels_images(term, PEXELS_API_KEY) if fallback_url: downloaded_fallback = download_image(fallback_url, fallback_file) if downloaded_fallback: print(f"Fallback image saved to {downloaded_fallback}") return {"path": downloaded_fallback, "asset_type": "image"} else: print(f"Fallback image download failed for term: {term}") else: print(f"Fallback image search failed for term: {term}") print(f"Failed to generate visual asset for prompt: {prompt}") return None def generate_silent_audio(duration, sample_rate=24000): """Generate a silent WAV audio file lasting 'duration' seconds.""" num_samples = int(duration * sample_rate) silence = np.zeros(num_samples, dtype=np.float32) silent_path = os.path.join(TEMP_FOLDER, f"silent_{int(time.time())}.wav") sf.write(silent_path, silence, sample_rate) print(f"Silent audio generated: {silent_path}") return silent_path def generate_tts(text, voice): """ Generate TTS audio using Kokoro, falling back to gTTS or silent audio if needed. """ safe_text = re.sub(r'[^\w\s-]', '', text[:10]).strip().replace(' ', '_') file_path = os.path.join(TEMP_FOLDER, f"tts_{safe_text}.wav") if os.path.exists(file_path): print(f"Using cached TTS for text '{text[:10]}...'") return file_path try: kokoro_voice = selected_voice if voice == 'en' else voice generator = pipeline(text, voice=kokoro_voice, speed=voice_speed, split_pattern=r'\n+') audio_segments = [] for i, (gs, ps, audio) in enumerate(generator): audio_segments.append(audio) full_audio = np.concatenate(audio_segments) if len(audio_segments) > 1 else audio_segments[0] sf.write(file_path, full_audio, 24000) print(f"TTS audio saved to {file_path} (Kokoro)") return file_path except Exception as e: print(f"Error with Kokoro TTS: {e}") try: print("Falling back to gTTS...") tts = gTTS(text=text, lang='en') mp3_path = os.path.join(TEMP_FOLDER, f"tts_{safe_text}.mp3") tts.save(mp3_path) audio = AudioSegment.from_mp3(mp3_path) audio.export(file_path, format="wav") os.remove(mp3_path) print(f"Fallback TTS saved to {file_path} (gTTS)") return file_path except Exception as fallback_error: print(f"Both TTS methods failed: {fallback_error}") return generate_silent_audio(duration=max(3, len(text.split()) * 0.5)) def apply_kenburns_effect(clip, target_resolution, effect_type=None): """Apply a smooth Ken Burns effect with a single movement pattern.""" target_w, target_h = target_resolution clip_aspect = clip.w / clip.h target_aspect = target_w / target_h if clip_aspect > target_aspect: new_height = target_h new_width = int(new_height * clip_aspect) else: new_width = target_w new_height = int(new_width / clip_aspect) clip = clip.resize(newsize=(new_width, new_height)) base_scale = 1.15 new_width = int(new_width * base_scale) new_height = int(new_height * base_scale) clip = clip.resize(newsize=(new_width, new_height)) max_offset_x = new_width - target_w max_offset_y = new_height - target_h available_effects = ["zoom-in", "zoom-out", "pan-left", "pan-right", "up-left"] if effect_type is None or effect_type == "random": effect_type = random.choice(available_effects) if effect_type == "zoom-in": start_zoom = 0.9 end_zoom = 1.1 start_center = (new_width / 2, new_height / 2) end_center = start_center elif effect_type == "zoom-out": start_zoom = 1.1 end_zoom = 0.9 start_center = (new_width / 2, new_height / 2) end_center = start_center elif effect_type == "pan-left": start_zoom = 1.0 end_zoom = 1.0 start_center = (max_offset_x + target_w / 2, (max_offset_y // 2) + target_h / 2) end_center = (target_w / 2, (max_offset_y // 2) + target_h / 2) elif effect_type == "pan-right": start_zoom = 1.0 end_zoom = 1.0 start_center = (target_w / 2, (max_offset_y // 2) + target_h / 2) end_center = (max_offset_x + target_w / 2, (max_offset_y // 2) + target_h / 2) elif effect_type == "up-left": start_zoom = 1.0 end_zoom = 1.0 start_center = (max_offset_x + target_w / 2, max_offset_y + target_h / 2) end_center = (target_w / 2, target_h / 2) else: raise ValueError(f"Unsupported effect_type: {effect_type}") def transform_frame(get_frame, t): frame = get_frame(t) ratio = t / clip.duration if clip.duration > 0 else 0 ratio = 0.5 - 0.5 * math.cos(math.pi * ratio) current_zoom = start_zoom + (end_zoom - start_zoom) * ratio crop_w = int(target_w / current_zoom) crop_h = int(target_h / current_zoom) current_center_x = start_center[0] + (end_center[0] - start_center[0]) * ratio current_center_y = start_center[1] + (end_center[1] - start_center[1]) * ratio min_center_x = crop_w / 2 max_center_x = new_width - crop_w / 2 min_center_y = crop_h / 2 max_center_y = new_height - crop_h / 2 current_center_x = max(min_center_x, min(current_center_x, max_center_x)) current_center_y = max(min_center_y, min(current_center_y, max_center_y)) cropped_frame = cv2.getRectSubPix(frame, (crop_w, crop_h), (current_center_x, current_center_y)) resized_frame = cv2.resize(cropped_frame, (target_w, target_h), interpolation=cv2.INTER_LANCZOS4) return resized_frame return clip.fl(transform_frame) def resize_to_fill(clip, target_resolution): """Resize and crop a clip to fill the target resolution while maintaining aspect ratio.""" target_w, target_h = target_resolution clip_aspect = clip.w / clip.h target_aspect = target_w / target_h if clip_aspect > target_aspect: clip = clip.resize(height=target_h) crop_amount = (clip.w - target_w) / 2 clip = clip.crop(x1=crop_amount, x2=clip.w - crop_amount, y1=0, y2=clip.h) else: clip = clip.resize(width=target_w) crop_amount = (clip.h - target_h) / 2 clip = clip.crop(x1=0, x2=clip.w, y1=crop_amount, y2=clip.h - crop_amount) return clip def find_mp3_files(): """Search for any MP3 files in the current directory and subdirectories.""" mp3_files = [] for root, dirs, files in os.walk('.'): for file in files: if file.endswith('.mp3'): mp3_path = os.path.join(root, file) mp3_files.append(mp3_path) print(f"Found MP3 file: {mp3_path}") return mp3_files[0] if mp3_files else None def add_background_music(final_video, bg_music_volume=0.10): """Add background music to the final video using any MP3 file found.""" try: bg_music_path = "music.mp3" if bg_music_path and os.path.exists(bg_music_path): print(f"Adding background music from: {bg_music_path}") bg_music = AudioFileClip(bg_music_path) if bg_music.duration < final_video.duration: loops_needed = math.ceil(final_video.duration / bg_music.duration) bg_segments = [bg_music] * loops_needed bg_music = concatenate_audioclips(bg_segments) bg_music = bg_music.subclip(0, final_video.duration) bg_music = bg_music.volumex(bg_music_volume) video_audio = final_video.audio mixed_audio = CompositeAudioClip([video_audio, bg_music]) final_video = final_video.set_audio(mixed_audio) print("Background music added successfully") else: print("No MP3 files found, skipping background music") return final_video except Exception as e: print(f"Error adding background music: {e}") print("Continuing without background music") return final_video def create_clip(media_path, asset_type, tts_path, duration=None, effects=None, narration_text=None, segment_index=0): """Create a video clip with synchronized subtitles and narration.""" try: print(f"Creating clip #{segment_index} with asset_type: {asset_type}, media_path: {media_path}") if not os.path.exists(media_path) or not os.path.exists(tts_path): print("Missing media or TTS file") return None audio_clip = AudioFileClip(tts_path).audio_fadeout(0.2) audio_duration = audio_clip.duration target_duration = audio_duration + 0.2 if asset_type == "video": clip = VideoFileClip(media_path) clip = resize_to_fill(clip, TARGET_RESOLUTION) if clip.duration < target_duration: clip = clip.loop(duration=target_duration) else: clip = clip.subclip(0, target_duration) elif asset_type == "image": img = Image.open(media_path) if img.mode != 'RGB': with tempfile.NamedTemporaryFile(suffix='.jpg', delete=False) as temp: img.convert('RGB').save(temp.name) media_path = temp.name img.close() clip = ImageClip(media_path).set_duration(target_duration) clip = apply_kenburns_effect(clip, TARGET_RESOLUTION) clip = clip.fadein(0.3).fadeout(0.3) else: return None if narration_text and CAPTION_COLOR != "transparent": try: words = narration_text.split() chunks = [] current_chunk = [] for word in words: current_chunk.append(word) if len(current_chunk) >= 5: chunks.append(' '.join(current_chunk)) current_chunk = [] if current_chunk: chunks.append(' '.join(current_chunk)) chunk_duration = audio_duration / len(chunks) subtitle_clips = [] subtitle_y_position = int(TARGET_RESOLUTION[1] * 0.70) for i, chunk_text in enumerate(chunks): start_time = i * chunk_duration end_time = (i + 1) * chunk_duration txt_clip = TextClip( chunk_text, fontsize=45, font='Arial-Bold', color=CAPTION_COLOR, bg_color='rgba(0, 0, 0, 0.25)', method='caption', align='center', stroke_width=2, stroke_color=CAPTION_COLOR, size=(TARGET_RESOLUTION[0] * 0.8, None) ).set_start(start_time).set_end(end_time) txt_clip = txt_clip.set_position(('center', subtitle_y_position)) subtitle_clips.append(txt_clip) clip = CompositeVideoClip([clip] + subtitle_clips) except Exception as sub_error: print(f"Subtitle error: {sub_error}") txt_clip = TextClip( narration_text, fontsize=font_size, color=CAPTION_COLOR, align='center', size=(TARGET_RESOLUTION[0] * 0.7, None) ).set_position(('center', int(TARGET_RESOLUTION[1] / 3))).set_duration(clip.duration) clip = CompositeVideoClip([clip, txt_clip]) clip = clip.set_audio(audio_clip) print(f"Clip created: {clip.duration:.1f}s") return clip except Exception as e: print(f"Error in create_clip: {str(e)}") return None def fix_imagemagick_policy(): """Fix ImageMagick security policies.""" try: print("Attempting to fix ImageMagick security policies...") policy_paths = [ "/etc/ImageMagick-6/policy.xml", "/etc/ImageMagick-7/policy.xml", "/etc/ImageMagick/policy.xml", "/usr/local/etc/ImageMagick-7/policy.xml" ] found_policy = next((path for path in policy_paths if os.path.exists(path)), None) if not found_policy: print("No policy.xml found. Using alternative subtitle method.") return False print(f"Modifying policy file at {found_policy}") os.system(f"sudo cp {found_policy} {found_policy}.bak") os.system(f"sudo sed -i 's/rights=\"none\"/rights=\"read|write\"/g' {found_policy}") os.system(f"sudo sed -i 's/]*>/]*>//g' {found_policy}") print("ImageMagick policies updated successfully.") return True except Exception as e: print(f"Error fixing policies: {e}") return False # ---------------- Main Video Generation Function ---------------- # def generate_video(user_input, resolution, caption_option): """Generate a video based on user input via Gradio.""" global TARGET_RESOLUTION, CAPTION_COLOR, TEMP_FOLDER # Set resolution if resolution == "Full": TARGET_RESOLUTION = (1920, 1080) elif resolution == "Short": TARGET_RESOLUTION = (1080, 1920) else: TARGET_RESOLUTION = (1920, 1080) # Default # Set caption color CAPTION_COLOR = "white" if caption_option == "Yes" else "transparent" # Create a unique temporary folder TEMP_FOLDER = tempfile.mkdtemp() # Fix ImageMagick policy fix_success = fix_imagemagick_policy() if not fix_success: print("Will use alternative methods if needed") print("Generating script from API...") script = generate_script(user_input) if not script: print("Failed to generate script.") shutil.rmtree(TEMP_FOLDER) return None print("Generated Script:\n", script) elements = parse_script(script) if not elements: print("Failed to parse script into elements.") shutil.rmtree(TEMP_FOLDER) return None print(f"Parsed {len(elements)//2} script segments.") paired_elements = [] for i in range(0, len(elements), 2): if i + 1 < len(elements): paired_elements.append((elements[i], elements[i + 1])) if not paired_elements: print("No valid script segments found.") shutil.rmtree(TEMP_FOLDER) return None clips = [] for idx, (media_elem, tts_elem) in enumerate(paired_elements): print(f"\nProcessing segment {idx+1}/{len(paired_elements)} with prompt: '{media_elem['prompt']}'") media_asset = generate_media(media_elem['prompt'], current_index=idx, total_segments=len(paired_elements)) if not media_asset: print(f"Skipping segment {idx+1} due to missing media asset.") continue tts_path = generate_tts(tts_elem['text'], tts_elem['voice']) if not tts_path: print(f"Skipping segment {idx+1} due to TTS generation failure.") continue clip = create_clip( media_path=media_asset['path'], asset_type=media_asset['asset_type'], tts_path=tts_path, duration=tts_elem['duration'], effects=media_elem.get('effects', 'fade-in'), narration_text=tts_elem['text'], segment_index=idx ) if clip: clips.append(clip) else: print(f"Clip creation failed for segment {idx+1}.") if not clips: print("No clips were successfully created.") shutil.rmtree(TEMP_FOLDER) return None print("\nConcatenating clips...") final_video = concatenate_videoclips(clips, method="compose") final_video = add_background_music(final_video, bg_music_volume=bg_music_volume) print(f"Exporting final video to {OUTPUT_VIDEO_FILENAME}...") final_video.write_videofile(OUTPUT_VIDEO_FILENAME, codec='libx264', fps=fps, preset=preset) print(f"Final video saved as {OUTPUT_VIDEO_FILENAME}") # Clean up print("Cleaning up temporary files...") shutil.rmtree(TEMP_FOLDER) print("Temporary files removed.") return OUTPUT_VIDEO_FILENAME # ---------------- Gradio Interface ---------------- # VOICE_CHOICES = { 'Emma (Female)': 'af_heart', 'Bella (Female)': 'af_bella', 'Nicole (Female)': 'af_nicole', 'Aoede (Female)': 'af_aoede', 'Kore (Female)': 'af_kore', 'Sarah (Female)': 'af_sarah', 'Nova (Female)': 'af_nova', 'Sky (Female)': 'af_sky', 'Alloy (Female)': 'af_alloy', 'Jessica (Female)': 'af_jessica', 'River (Female)': 'af_river', 'Michael (Male)': 'am_michael', 'Fenrir (Male)': 'am_fenrir', 'Puck (Male)': 'am_puck', 'Echo (Male)': 'am_echo', 'Eric (Male)': 'am_eric', 'Liam (Male)': 'am_liam', 'Onyx (Male)': 'am_onyx', 'Santa (Male)': 'am_santa', 'Adam (Male)': 'am_adam', 'Emma 🇬🇧 (Female)': 'bf_emma', 'Isabella 🇬🇧 (Female)': 'bf_isabella', 'Alice 🇬🇧 (Female)': 'bf_alice', 'Lily 🇬🇧 (Female)': 'bf_lily', 'George 🇬🇧 (Male)': 'bm_george', 'Fable 🇬🇧 (Male)': 'bm_fable', 'Lewis 🇬🇧 (Male)': 'bm_lewis', 'Daniel 🇬🇧 (Male)': 'bm_daniel' } def generate_video_with_options(user_input, resolution, caption_option, music_file, voice, vclip_prob, bg_vol, video_fps, video_preset, v_speed, caption_size): global selected_voice, voice_speed, font_size, video_clip_probability, bg_music_volume, fps, preset # Update global variables with user selections selected_voice = VOICE_CHOICES[voice] voice_speed = v_speed font_size = caption_size video_clip_probability = vclip_prob / 100 # Convert from percentage to decimal bg_music_volume = bg_vol fps = video_fps preset = video_preset # Handle music upload if music_file is not None: target_path = "music.mp3" shutil.copy(music_file.name, target_path) print(f"Uploaded music saved as: {target_path}") # Generate the video return generate_video(user_input, resolution, caption_option) # Create the Gradio interface iface = gr.Interface( fn=generate_video_with_options, inputs=[ gr.Textbox(label="Video Concept", placeholder="Enter your video concept here..."), gr.Radio(["Full", "Short"], label="Resolution", value="Full"), gr.Radio(["Yes", "No"], label="Captions", value="Yes"), gr.File(label="Upload Background Music (MP3)", file_types=[".mp3"]), gr.Dropdown(choices=list(VOICE_CHOICES.keys()), label="Choose Voice", value="Emma (Female)"), gr.Slider(0, 100, value=25, step=1, label="Video Clip Usage Probability (%)"), gr.Slider(0.0, 1.0, value=0.08, step=0.01, label="Background Music Volume"), gr.Slider(10, 60, value=30, step=1, label="Video FPS"), gr.Dropdown(choices=["ultrafast", "superfast", "veryfast", "faster", "fast", "medium", "slow"], value="veryfast", label="Export Preset"), gr.Slider(0.5, 1.5, value=0.9, step=0.05, label="Voice Speed"), gr.Slider(20, 100, value=45, step=1, label="Caption Font Size") ], outputs=gr.Video(label="Generated Video"), title="AI Documentary Video Generator", description="Create short documentary videos with AI. Upload music, choose voice, and customize settings." ) # Launch the interface if __name__ == "__main__": iface.launch(share=True)