I Built a Full YouTube Automation Tool With Cursor AI (Auto Upload, Shorts, Descriptions & More)
Posting a single YouTube video used to eat up an hour of my day — writing the description, crafting a first comment, cutting shorts, uploading everything by hand. So I built a tool to do it for me. In this post I'm breaking down the YouTube automation tool I built inside Cursor, what it actually does, and how the whole pipeline fits together.
What the Tool Does
At its core, this is a Python-based automation pipeline that takes a video from raw file to fully published YouTube upload with almost no manual input. It started as a simple description generator and has grown into a full content pipeline that handles transcripts, descriptions, comments, uploads, and now automatic Shorts creation.
Why This Matters
If you post consistently on YouTube, you already know the repetitive work adds up fast: writing SEO-friendly descriptions, dropping a first comment, cutting a long video into shorts, writing separate descriptions for each of those, and uploading it all. None of that is hard, it's just slow. Automating it means more time actually making content and less time on busywork.
How the Build Works
The tool is built around a few key pieces working together:
- Transcription: faster-whisper generates a transcript directly from the local video file.
- AI content generation: Claude takes that transcript and writes an SEO-friendly description and a first comment automatically.
- Upload automation: the YouTube Data API handles pushing the video live without manual uploading.
- Shorts generation: the pipeline slices the long-form video into multiple Shorts using FFmpeg and FFprobe, cropping and formatting each clip with blurred backgrounds.
- Shorts metadata: each Short gets its own AI-generated description, comment, and thumbnail, then gets posted automatically.
I also built a GUI on top of the Python script so I'm not running everything from the command line, plus a video management dashboard where I can pull up existing YouTube videos, edit titles, swap thumbnails, and regenerate content after the fact. There's a blog integration piece in progress too, aimed at auto-publishing videos to dedicated pages on this site.
The Hard Part: Smart Cropping
The trickiest challenge so far has been getting the tool to intelligently follow the action on screen when cropping long-form video down into a vertical Shorts format. I go through what worked, what didn't, and what I ended up scrapping for now in the walkthrough below.
Key Takeaways
- One tool handles transcript generation, description writing, comment creation, uploading, and Shorts creation
- Built with Cursor, Claude, faster-whisper, FFmpeg/FFprobe, and the YouTube Data API
- Includes a GUI and a video management dashboard, not just a script
- Blog auto-publishing is in active development
- Turns roughly an hour of manual posting work into a hands-free process
Want to see the actual build process, including the Cursor prompts, the Python script, and the tool running end to end? Click the button below.
If you're building your own workflow tools or just want to follow along as this one evolves, subscribe and stick around — more builds like this are on the way.