Faceless Reel Generator (Thrice AI)

An advanced, fully automated tool to generate high-quality vertical video reels. By leveraging AI for scriptwriting, multi-engine text-to-speech, and automated video compilation (complete with subtitles, avatars, and PIP features), this project turns simple topics or YouTube RSS feeds into engaging content. You can interact with it via a local Next.js Web App dashboard or entirely through a secure Telegram Bot.

Key Features

• AI Script Generation: Automatically drafts contextual dialogue using Gemini AI.
• Multiple TTS Engines Supported:
  • ElevenLabs: High-quality, premium cloud voice cloning.
  • Gemini TTS: Integrated cloud TTS from Google.
  • Kokoro: Local, fast, ONNX-based open-source TTS.
  • macOS Native (say): Built-in, zero-dependency offline TTS for Mac users.
• Rich Video Compilation (MoviePy / FFmpeg):
  • Dynamic overlay of animated subtitles with customizable fonts and stroke colors.
  • Automated integration of Avatar images (bouncing animation when speaking).
  • Picture-in-Picture (PIP) asset overlay support.
  • Random stock video background selection and background music integration.
• Telegram Bot Interface: Control the entire pipeline securely from your phone.
• Content Ingestion: Capable of digesting YouTube RSS feeds for trending topics.

Project Architecture & Workflow

graph TD
    A[User Input / YouTube RSS Feed] --> B[Gemini AI: Script Generation]
    B --> C[TTS: ElevenLabs / Kokoro / Gemini / mac_say]
    B --> D[Stock Background Videos / Gameplay]
    B --> I[Avatar & PIP Asset Selection]
    C --> E[Video Compiler: MoviePy / FFmpeg]
    D --> E
    I --> E
    E --> F[Subtitles / Captions Overlay]
    F --> G[Background Music Integration]
    G --> H[Final Generated Reel .mp4]

Loading

Prerequisites

• Python: Version 3.12 or higher.
• Package Manager: uv (Recommended) or pip.
• System Tools:
  • FFmpeg: Essential for video processing.
  • ImageMagick: Required by MoviePy for text/subtitle generations.

Installation

macOS

1. Install System Dependencies (using Homebrew):

   brew install ffmpeg imagemagick

   Note: Ensure the MAGICK_HOME environment variable is set or convert/magick is in your PATH.

2. Install uv (Recommended):

   curl -LsSf https://astral.sh/uv/install.sh | sh

Linux / Windows

Install Python, FFmpeg, and ImageMagick via your OS package manager (apt, dnf, choco, etc.).

Project Setup

1. Clone the Repository:

   git clone <repository_url>
   cd thriceai

2. Install Python Dependencies:

   uv sync
   # OR: pip install -e .

3. Setup Frontend (Optional for Dashboard):

   cd web_app
   bun install

4. Prepare Local Assets (Required Directories): Ensure you place your local models and media in the data/ directory:

   • data/assets/bg_videos/: Drop your background MP4s here.
   • data/assets/avatars/: PNG avatars for characters.
   • data/models/kokoro/: (If using Kokoro) Place kokoro-v1.0.onnx and voices-v1.0.bin here.
   • data/fonts/: Place your Inter_28pt-ExtraBold.ttf or preferred fonts.

Configuration (.env)

Create a .env file in the root directory. Configure only what you need:

# --- LLM API Keys ---
GEMINI_API_KEY=your_gemini_api_key_here
CLAUDE=your_claude_api_key_here

# --- TTS Providers ---
ELEVEN_API=your_elevenlabs_api_key_here

# --- Telegram Bot Configuration ---
TELEGRAM_BOT=your_telegram_bot_token_here
TELEGRAM_CHAT_ID=@your_channel_or_chat_id
# CRITICAL: Comma-separated Telegram User IDs allowed to use bot commands
AUTHORIZED_TELEGRAM_USERS=123456789,987654321

Usage

1. Telegram Bot (Recommended)

The bot provides a secure, remote interface for the generation pipeline:

1. Start the backend: uv run run.py
2. Message your bot on Telegram:
   • /help: List all commands.
   • /script <topic> <filename>: Generate a script.
   • /generate <filename> [audio_mode]: Generate a reel from a script. Supported audio modes: kokoro_mlx, kokoro, mac_say, elevenlabs, gemini.
   • /reel <topic>: Full Auto Mode (Script -> Reel -> Caption -> Telegram Upload).
   • /status: Check generation progress.

2. Local Web Dashboard

For local visual management:

1. Start the Backend: uv run run.py (Hosts API on Port 8008)
2. Start the Frontend:

   cd web_app
   bun run dev

3. Open http://localhost:3031 (or local IP) in your browser.

Screenshots

Main Dashboard & Generation Workflow

Initial view showing feed and batch generation modes.

Script Generation

Input interface for script generation.

The system generating a script based on user input.

Video Generation & Monitoring

Batch generation of reels.

Real-time updates during the generation process.

Monitoring individual video generation tasks.

Components & Insights

Adding visual components like characters and PIP.

Post-generation insights and analytics.

Troubleshooting & Notes

• ImageMagick Policy Error: If ImageMagick blocks @ symbols (needed for text files), edit /etc/ImageMagick-6/policy.xml (location varies) to allow reading .txt files.
• Missing Fonts: Ensure the font paths in backend/config.py correctly point to existing .ttf files in your data/fonts/ directory.