From Prompt to Reel: Automating Vertical Video Creation

Let's be honest: creating short-form vertical video content is a grind. You have the idea, maybe even a script, but then you're stuck storyboarding, finding clips, editing, adding captions, and fitting it all into a tight 9:16 frame. What if you could skip straight from the idea to a finished draft?

That's the itch that the open-source project YumCut scratches. It takes a simple text prompt and automatically generates a structured, edited vertical video. It's like having a first-pass video editor that works at the speed of your typing.

What It Does

YumCut is a web application that automates the initial, most time-consuming stages of vertical video creation. You give it a topic—like "benefits of morning sunlight"—and it goes to work. Using AI, it generates a short script, sources relevant stock video clips for each scene, stitches them together in sequence, and overlays auto-generated captions that match the pacing of the voiceover. The output is a ready-to-review MP4 file, formatted perfectly for platforms like TikTok, Instagram Reels, or YouTube Shorts.

Why It's Cool

The clever part isn't just that it uses AI models; it's how it orchestrates them into a coherent pipeline. Instead of being a single, monolithic model, YumCut breaks down the video creation process into discrete, manageable steps: ideation, asset sourcing, assembly, and publishing. This modular approach is not only more reliable but also transparent and, importantly, hackable for developers.

For creators, it's a powerful brainstorming and prototyping tool. You can validate a video concept in minutes instead of hours. For developers, the open-source code on GitHub is a fascinating look at a modern, full-stack AI application. It's built with Next.js, uses OpenAI for script generation, leverages search APIs for video clips, and employs FFmpeg for the heavy lifting of video processing—a great real-world example of several technologies working in concert.

How to Try It

The easiest way to see it in action is to use the live demo. Just head over to the project's hosted version:

Try YumCut Live

On the site, you'll find a simple input field. Type in your video topic, hit generate, and wait a couple of minutes while the magic happens. You can then preview and download your video.

If you're the hands-on type and want to run it locally or see how it's built, the entire codebase is available on GitHub. The README provides setup instructions if you have your own API keys for the required services.

git clone https://github.com/IgorShadurin/app.yumcut.com

Final Thoughts

YumCut won't (and shouldn't) replace final human editing for polished content. What it does exceptionally well is eliminate the blank canvas problem. It gets you to a solid first draft instantly, which you can then tweak, refine, or use as inspiration. For developers, it's a compelling blueprint for building practical, multi-step AI applications. Whether you use it to quickly create content or learn from its architecture, this project is a neat example of using automation to handle the boring parts so you can focus on the creative ones.

Follow us for more interesting projects: @githubprojects