Gemini YouTube Automation: Fully Automated AI Video Creation and Publishing System

Original Author/Maintainer: ChaitanyaEswarRajeshJakki Source Platform: GitHub Original Link: https://github.com/ChaitanyaEswarRajeshJakki/gemini-youtube-automation

Gemini YouTube Automation is a fully autonomous AI agent project developed in Python. It uses the Google Gemini 2.5 Flash large language model and GitHub Actions to implement fully automated scheduled tasks, enabling daily generation of educational videos, rendering of professional videos, and uploading to YouTube without any human intervention. This project is an end-to-end automation pipeline covering the entire process from content planning to publishing. Currently, it has 287 Stars and 136 Forks, with high community attention.

At a time when AI is reshaping the content creation field, Gemini YouTube Automation is positioned as a complete end-to-end automation pipeline. It is not just a video generation tool but also integrates all links from content planning, script writing, audio synthesis, visual rendering to publishing. The project is producing the "AI for Developers" series of courses, covering multiple topics such as generative AI, large language models, and prompt engineering, to solve the problem of educational content production efficiency.

The system is triggered daily at UTC 7:00, and the workflow is as follows:

1. Read content_plan.json to select the course topic to be processed;
2. Call Gemini 2.5 Flash to generate scripts for 7-8 slides, Shorts summaries, tags, and metadata;
3. Use gTTS to convert the script into voice narration, and Pexels API to obtain matching background images;
4. Use MoviePy + FFmpeg to render landscape (1920×1080) and portrait (1080×1920) videos;
5. Mix the narration with background music and generate custom thumbnails;
6. Upload the video via YouTube Data API v3, update content_plan to mark it as completed, and commit to the repository.

• AI Content Generation: Google Gemini 2.5 Flash (balances quality and cost, long context + structured output);
• Speech Synthesis: gTTS (free and lightweight, suitable for automation scenarios);
• Video Rendering: MoviePy + FFmpeg (fine control over video/audio, supports dual-format output);
• Visual Materials: Pexels API (high-quality CC0 images);
• Infrastructure: GitHub Actions (zero server cost, scheduled triggering + Secrets management for sensitive information).

Deployment Steps:

1. Clone the repository and install dependencies (moviepy, gTTS, Pillow, etc.);
2. Set up GitHub Secrets: Google API Key, Pexels API Key, Base64-encoded OAuth credentials (client_secrets.json, credentials.json);
3. Local Testing: Run python main.py;
4. Production Run: GitHub Actions is automatically triggered daily at UTC 7:00;
5. Quota Handling: Only upload 1 long video + 1 Shorts per cycle to avoid YouTube API rate limits.

Application Scenarios: Automated operation for educational content creators (technical lecturers, online education institutions, personal IPs to reduce marginal production costs); Deep Value: Demonstrates end-to-end automation of AI agents across modalities (text → speech → video) and platforms (GitHub → YouTube); Expansion Directions: Replace the generation model (Claude/GPT-4), upgrade TTS (ElevenLabs/Azure), use AI image generation instead of image libraries, migrate to private servers/K8s clusters.

This project represents a new paradigm of content creation: "human creativity + AI execution", freeing creators from repetitive technical work to focus on course architecture and knowledge organization; For the technical community: Provides a reference implementation for AI engineering (clear code, complete documentation); Future Outlook: Advances in multimodal AI will make automation tools more intelligent and popular, and creators will increasingly play the roles of "curators" and "directors".