🎬 AutoShorts AI: The Automated Faceless Video Generator

AutoShorts AI is a Python pipeline that creates viral-style "Faceless" YouTube Shorts and TikToks from a topic. It handles the production chain: AI topic/script generation, voiceover generation, stock footage sourcing, and FFmpeg editing with transitions and avatar injection.

✨ Key Features

• 🧠 Intelligent Scriptwriting: Uses Google Gemini 2.0 Flash to write engaging, "Edutainment" style scripts (Vox/Kurzgesagt style) with strict storytelling structures (Hook → Context → Mechanism → Twist).
• 🗣️ Voiceovers: Generates narration with edge-tts.
• 🎞️ Dual-Visual System: Automatically searches and downloads two distinct stock videos per scene from Pexels, creating a dynamic "A/B Split" visual style to maximize viewer retention.
• ✂️ Advanced FFmpeg Editing:
• Smart Trimming: Syncs video perfectly to audio duration.
• A/B Splitting: Cuts every scene in half, switching visuals mid-sentence.
• Pro Transitions: Randomly applies xfade (fade, slide, wipes) between scenes.
• Silence Removal: Automatically trims dead air from AI voice generation.

• 🤖 Random Avatar Injection: Automatically inserts a custom "Avatar/Mascot" video into a random middle scene to build channel brand identity.
• 🪟 Windows Ready: Includes specific FFmpeg flags (yuv420p, faststart) to prevent corruption errors (0x80004005) on Windows Media Player.

📂 Project Structure

Automated-YT-Shorts-AI/
│
├── assets/                  # Stores all media files
│   ├── audio_clips/         # Generated voiceovers (.wav)
│   ├── video_clips/         # Downloaded stock footage (.mp4)
│   ├── temp/                # Intermediate processing files
│   ├── final/               # 🏆 The Final Output Video lives here
│   └── avatar/              # ⚠️ PUT YOUR AVATAR VIDEO HERE
│       └── ProfessionalGirlAnimationVideoGeneration.mp4
│
├── modules/                 # Core Logic Modules
│   ├── brain.py             # AI Scriptwriter (Gemini)
│   ├── audio.py             # Voice generator (edge-tts)
│   ├── asset_manager.py     # Pexels Downloader (Dual-Visual logic)
│   └── composer.py          # FFmpeg Video Editor (Stitching & Transitions)
│
├── main.py                  # Entry point (Orchestrator)
└── requirements.txt         # Python dependencies

🛠️ Prerequisites

• Python 3.10+ installed.
• FFmpeg installed and added to your system PATH.

• Windows: winget install ffmpeg (or download from ffmpeg.org).
• Verify: Type ffmpeg -version in your terminal.

• API Keys:

• Google Gemini API Key (Free tier available).
• Pexels API Key (Free).
• No Ngrok token is required for the default voiceover path. The current pipeline uses edge-tts.

🚀 Installation

1. Clone the Repository

git clone https://github.com/yourusername/AutoShorts-AI.git
cd AutoShorts-AI

2. Install Dependencies

pip install -r requirements.txt

3. Environment Setup

Create the required folders and add your avatar:

• Create folder: assets/avatar
• Place your avatar video inside and name it: avatars.mp4

4. Configure API Keys

Copy .env.example to .env and fill in your key:

cp .env.example .env

Required:

• GEMINIAPIKEY for script generation
• PEXELSAPIKEY for stock video search/download

• GEMINI_MODEL to override the default gemini-2.0-flash model

🎮 How to Run

Generate Video

Run the main script:

python main.py

• Enter a topic (e.g., "The Mystery of the Pyramids").
• Wait for the AI to write the script, generate audio, download stock footage, and edit the video.
• The final video will be saved in assets/final/final_short.mp4.

🧩 Module Breakdown

brain.py ( The Writer)

• Input: Topic string.
• Logic: Prompts Gemini to create an 8-9 scene JSON script. It asks for two visual keywords per scene (visual1, visual2) to enable the A/B split effect.

audio.py (The Voice)

• Input: Text script.
• Logic: Generates MP3 voice clips with edge-tts.
• Post-Processing: Reads durations with mutagen so scenes can be synced to audio length.

asset_manager.py (The Librarian)

• Input: Visual keywords.
• Logic: Searches Pexels for Portrait (9:16) videos. Downloads pairs of videos for every scene. Handles fallbacks (if Video B is missing, reuse Video A).

composer.py (The Editor)

• Input: Audio files + Video files.
• Logic:
• Scene Processing: Cuts the scene duration in half. Plays Video A for the first half, Video B for the second half.
• Avatar Injection: Identifies a random "middle" scene (not hook/outro) and replaces the stock footage with your Avatar loop.
• Stitching: Merges all scenes using xfade transitions (wipes, slides).
• Rendering: Exports as yuv420p H.264 MP4 with faststart flags for maximum compatibility.

⚠️ Troubleshooting

Q: The video is black or corrupt (0x80004005 error).

• Fix: This is usually a Windows codec issue. The updated composer.py forces pix_fmt='yuv420p'. Try opening the file with VLC Media Player.

• Fix: Ensure your folder structure is exactly assets/avatar/avatars.mp4.

• Fix: Check your internet connection and that edge-tts is installed from requirements.txt.

• Fix: Ensure FFmpeg is installed and accessible from your command line.

📜 License

This project is open-source. Feel free to modify and build your own automation empire!