Viral Faceless Shorts Generator

An automated pipeline to generate faceless YouTube Shorts based on Google Trends or custom topic. This system leverages AI-generated scripts, manual script approval, TTS voice, subtitle alignment, and FFmpeg video composition, all wrapped in a fully containerized architecture.

❤️ Support This Project

Maintaining and improving this open-source project takes significant time and effort. If you find it useful, please consider supporting my work via GitHub Sponsors:

👉 GitHub Sponsors

Your support helps keep this project alive and free for the community. Thank you! 🙌

Features

• Trending Topic Detection via Google Trends
• Prompt-based script generation with Google Gemini
• Manual script approval and editing before rendering
• Voice synthesis with Piper TTS
• Forced alignment of subtitles with Aeneas
• Video composition with FFmpeg
• Selectable output orientation (portrait, landscape, square, original, custom)
• Persistent SQLite-backed app settings (/api/settings)
• Configurable generated video save folder (server-side)
• One-click web interface

Pipeline Overview

1. Write Prompt/Choose a topic Choose a category to filter for trending topics from Google Trends, or choose a custom topic or enter a custom prompt in the web UI to describe the video idea.

2. Generate Script (Gemini) The trendscraper service uses Gemini to generate a title, description, and body.

3. Approve Script Review and optionally edit the script. Rendering starts only after approval.

4. Text-to-Speech (Piper) The approved script is converted to speech using Piper TTS.

5. Subtitle Alignment (Aeneas) speechalign aligns the generated audio with the script to produce timed subtitles.

6. Video Assembly (FFmpeg) The system merges:

   • A background video from /videos (named default_*.mp4)
   • The voiceover audio
   • Subtitles to create the final faceless short.

7. Web Trigger Interface Access a minimal frontend on port 80 to trigger everything with one click.

Folder Structure

viral_faceless/
├── trendscraper/      # Gemini + FFmpeg composition API
├── piper/             # Piper TTS container
├── speechalign/       # Forced alignment via Aeneas
├── nginx/             # Web interface + reverse proxy
├── videos/            # Background video clips (e.g., gameplay, press machine)
├── docker-compose.yml

Note: Background clips must be placed in /videos and named like default_something.mp4.

Usage

Rename example.env in .env and replace the placeholder Gemini API key with the one you generated from Google AI Studio, then start the pipeline with:

docker-compose up --build

Access the web interface at http://localhost to generate your first viral Short.

Requirements

• Docker + Docker Compose
• Google Gemini API Key
• At least one video clip in /videos/