An individual open-sourced a system of 10 AI agents that automate YouTube Shorts creation from podcast clips, including transcription, editing, captioning, and scheduling, with a quality-checking agent.
I have no audience and I am not a professional developer. I wanted a channel that could run itself, so I started building one agent, then it needed another agent to check its work, and about a month later I had ten of them. Here is what it actually does. You give it a YouTube podcast link. It transcribes the episode locally, scores which moments are most likely to perform, cuts them to vertical 9:16 with face tracking, burns in captions, adds music and effects, then checks its own output and schedules the posts. The agents: - **Finder** transcribes and scores clip-worthy moments - **Editor** cuts, reframes, captions, adds music and zooms - **Uploader** writes the titles and hashtags, posts to YouTube Shorts - **Manager** reads your real metrics and feeds what wins back into the Finder - **Trend Scout** checks what is trending in your niche - **Planner** decides the creative direction per clip - **Community** drafts comment replies - **Finishing Editor** watches every finished render and blocks broken ones (captions covering a face, dead air, frozen frames, bad audio) - **Trainer** studies top performers weekly and updates the playbooks - **Compiler** stitches the week's best moments into a long-form episode The part I am most happy with is the Finishing Editor. Everything else generates, but that one is the only agent whose whole job is to say "no, that one is broken, do not post it." It caught more bad clips than I expected. It runs on free AI providers (OpenRouter, Groq, Gemini), transcription is local with faster-whisper, and the music is synthesized so it is safe to monetize. So the running cost is basically zero. It is MIT licensed. Use it, change it, sell whatever you build with it. Honestly, if someone takes this and actually makes money with it, that would make my month. Fair warning: it is not perfect. The output quality varies, some parts are held together with duct tape, and I am sure there are bugs I have not hit yet. I am putting it out as it is rather than polishing forever. If people want to help fix it, that would be great. There is a PDF setup guide in there. Install is a double click on Windows, one command on Mac and Linux, then a wizard asks you a few questions. Happy to answer anything.
An open-source repository called 'The Agency' with 147 specialized AI agents across 12 divisions has rapidly gained 50K GitHub stars, representing a structured approach to AI-assisted development.
The author built an AI agent that pairs a YouTube MCP server with a CreatorLens skill to act as a growth strategist, diagnosing why videos perform and spotting outlier channels.
An open-source Python script automates creating faceless YouTube Shorts and TikTok videos from Reddit threads, using AI voiceover and automatic overlay, enabling easy content creation with zero cost.