YouTubeShortsBot

An automated, multi-platform bot that generates short‑story videos with AI and publishes them on a schedule to YouTube, Facebook, and Instagram.

The bot generates original horror/mystery stories with Google Gemini, turns them into vertical (9:16) videos with on‑screen text and narration, and uploads them automatically at configured times throughout the day. It supports two formats:

• Shorts – ~20s clips with the story text overlaid on a random background video and music.
• Long videos – ~5‑minute narrated stories built with FFmpeg, using Google Cloud Text‑to‑Speech for voiceover.

Features

• 🤖 AI story generation via Google Gemini (gemini-2.5-flash) across a wide range of horror/mystery categories, with built‑in fallback stories if the API is unavailable.
• 🎬 Short video creation using MoviePy: random background video + audio track + title/story text overlay, rendered to 1080×1920.
• 🎙️ Long video creation using FFmpeg with Google Cloud TTS narration and background music mixing.
• 📤 Multi-platform publishing to YouTube (resumable upload), Facebook Reels (chunked upload), and Instagram (stub — see notes).
• ⏰ Scheduling of uploads at configurable times (IST), with automatic Facebook long‑lived token refresh every 55 days.
• 🐳 Docker support with FFmpeg bundled in the image.

Project structure

.
├── main.py                    # Entry point: scheduling, orchestration, token refresh
├── story_generator.py         # Gemini story generation (short & long) + fallbacks
├── video_creator.py           # Short videos (MoviePy) + long video orchestration & Google TTS
├── ffmpeg_video_creator.py    # Long video rendering via raw FFmpeg
├── platform_uploaders.py      # YouTube / Facebook / Instagram uploaders
├── config/
│   └── constants.py           # Paths, scopes, fallback stories
├── assets/
│   ├── background_videos/     # Background clips (.mp4/.avi/.mov/.mkv)
│   ├── audio_tracks/          # Background music (.mp3/.wav/.aac)
│   └── fonts/                 # TTF fonts used for text overlays
├── data/                      # Generated output (videos, audio) — gitignored
├── requirements.txt
├── Dockerfile
└── Jenkinsfile

Requirements

• Python 3.12+
• FFmpeg and ffprobe available on PATH (required for long videos)
• Python dependencies in requirements.txt: moviepy, google-auth, google-api-python-client, google-auth-oauthlib, requests, pytz, schedule, python-dotenv, google-genai, google-cloud-texttospeech, google-cloud-storage

Setup

1. Clone and install dependencies

   git clone https://github.com/sumannath/YouTubeShortsBot.git
   cd YouTubeShortsBot
   pip install -r requirements.txt

2. Add assets to the assets/ directory:

   • One or more background videos in assets/background_videos/
   • One or more background music tracks in assets/audio_tracks/
   • A TTF font in assets/fonts/ (referenced by FONT_PATH)

3. Configure credentials in config/:

   • config/.env — environment variables (see below)
   • config/client_secret.json — Google OAuth client secrets for the YouTube Data API. On first run a browser flow generates config/yt_token.json.

4. Google Cloud (required for long videos): enable the Text‑to‑Speech and Cloud Storage APIs, create a bucket for synthesized audio, and authenticate (e.g. set GOOGLE_APPLICATION_CREDENTIALS to a service‑account key).

Configuration

Create config/.env with the variables your enabled platforms and formats need.

Core

Story generation

Google Cloud TTS / Storage (long videos)

Facebook

Instagram

Note: The Instagram uploader is currently a placeholder/stub and does not yet perform real uploads.

Usage

python main.py

On startup the bot:

1. Loads config/.env and validates required variables.
2. Runs an initial Facebook token refresh and generates/uploads one long video (useful for testing).
3. Enters the scheduler loop, publishing shorts and long videos at the configured *_UPLOAD_TIMES and refreshing the Facebook token every 55 days at 03:00.

Successfully uploaded local video files are deleted automatically; the loop continues running until interrupted (Ctrl+C).

Docker

docker build -t youtubeshortsbot .
docker run --rm \
  -v "$(pwd)/config:/app/config" \
  -v "$(pwd)/assets:/app/assets" \
  -v "$(pwd)/data:/app/data" \
  youtubeshortsbot

The image is based on python:3.12-slim and installs FFmpeg and required system libraries.

License

No license file is currently included. Add one if you intend others to reuse this code.