AWS Builder Center
Weekend Annoying Task Challenge: I Automated the Most Boring Part of Making YouTube Videos

Weekend Annoying Task Challenge: I Automated the Most Boring Part of Making YouTube Videos

I got tired of manually timestamping every YouTube video I made, so I built an AI-powered chapter creator in 2 hours using AWS Kiro and deployed it on AWS Amplify. It combines Google Gemini's video understanding with real-time YouTube validation so you never paste broken chapters again.

Software Engineer||Youtuber||Mentor
After editing a tutorial, I'd have to watch the entire video again, note every timestamp manually, write chapter titles, and make sure everything followed YouTube's strict chapter requirements. It was repetitive, slow, and easy to get wrong.
The obvious shortcut is to just ask ChatGPT or Gemini directly. Paste the URL, ask for chapters, done right?
Not quite.
What you get back is a block of text in a chat window. You still have to manually copy it, open YouTube Studio in another tab, paste it into the description, then discover that the first chapter is at 00:15 instead of 00:00 so YouTube silently ignores all your chapters. Or two chapters are 8 seconds apart and YouTube drops them without telling you. Back to the chat, ask it to fix specific lines, copy again, paste again. Every step is manual and there's zero feedback on what's actually broken.
The real problem isn't generating the timestamps it's the workflow around them. Verifying them, fixing them, validating them against YouTube's rules, and getting them into the right format.
So I built an AI-powered YouTube Chapter Creator paste a URL, and the entire workflow is handled in one place.
Why This Is Different From Just Using ChatGPT Here's the honest comparison:

ChatGPT gives you text. This gives you a workflow. ChatGPT is great for generating a first draft. This app is for finishing the job verifying, correcting, validating, and exporting in one place without switching tabs.
Two Ways to Work
Manual Mode — watch at up to 2.5× speed, hit a hotkey when a new topic starts, type the title. Keyboard shortcuts mean your hands never leave the keyboard.
AI Mode — click AI Draft, and Google Gemini watches the video and generates timestamped chapter suggestions. Review, edit, merge or replace before applying anything.

Every Feature

Embedded YouTube player with live timecode and playback speed controls (1× to 2.5×) Keyboard shortcuts: Alt+T or Spacebar captures timestamp, J/L or arrow keys seek ±10 seconds Timestamp list with inline title editing, ±1s fine-tuning, and segment duration badges Real-time YouTube chapter validation — checks all three of YouTube's rules (first chapter must be 00:00, minimum 3 chapters, chapters must be 10+ seconds apart) with auto-fix options AI draft review modal — uncheck entries, edit titles inline, choose merge or replace, then apply Export in YouTube Studio format (00:00 - Title) or Markdown clickable links ( 00:00 ) Copy button locks automatically if validation errors exist — you physically can't paste broken chapters into YouTube Studio Auto-save to localStorage — close the tab, come back tomorrow, everything is still there Responsive layout — two-column on desktop, stacked on mobile

How I Built It I built the entire application in about 2 hours using AWS Kiro and deployed it to AWS Amplify.

The core loop came together quickly — embed the YouTube IFrame Player API, capture timestamps on keypress, render an editable list. The trickier parts were:
YouTube validation logic — YouTube silently ignores chapters if you break any of their three rules. Most chapter tools skip this entirely. I built a real-time validation engine that checks all three rules on every keystroke and blocks the copy button on errors — so you can't accidentally paste broken chapters.
The AI draft system — A naive Gemini API call only analyzed the first few minutes of a video. The fix was calculating anchor timestamps based on video duration, embedding them in the prompt, and telling Gemini exactly how many chapters to generate and that they must span the full video. I also added a three-tier fallback: multimodal call → text-only call → evenly-spaced placeholder scaffold. A model cascade (gemini-2.5-flash-lite → gemini-2.5-flash → gemini-2.0-flash) handles Google's frequent model deprecations automatically — if one model name gets retired, it tries the next one without breaking.
Persistence without a backend — Everything saves to localStorage on every change. No account, no server, no save button needed.
AWS Architecture AWS Amplify — hosting and continuous deployment. Connected to GitHub; every push triggers an automatic build and deploy.
AWS Kiro — used to scaffold and build the application itself. Dramatically cut development time from a full day to a couple of hours.
GitHub Repo │ ▼ AWS Amplify (Build + Host) │ ▼ Static React App (Browser) ├── YouTube IFrame Player API ├── Google Gemini API └── localStorage (persistence)
No backend, no database. Pure static Vite build deployed straight to Amplify. The Gemini API key is injected at build time via Amplify's environment variables panel — a Lambda proxy is the planned next step for fully public sharing.

What I Learned

AWS Kiro dramatically speeds up scaffolding — spec docs, hooks, components, and wiring all generated and working in one session YouTube IFrame Player API has quirks: player state polling, event timing, and the player mount lifecycle all need careful handling that the docs underplay Prompt engineering matters a lot — vague prompts to Gemini produced 2-3 generic chapters; precise duration-aware prompts with anchor timestamps produced consistent full-video coverage Client-side validation as UX — locking the copy button is more effective than a warning banner because it makes the broken state impossible to ignore AWS Amplify makes static deployment nearly frictionless — CI/CD from GitHub live in minutes, and environment variables are managed cleanly without touching code
What's Next Serverless Lambda proxy to keep the Gemini API key server-side for public sharing Import existing chapters from published YouTube videos via the YouTube Data API Waveform visualization to spot topic transitions visually Multi-language support for non-English videos

Try It

 🔗 Live App: https://master.d11sd2swsvajzm.amplifyapp.com/  

Run it locally:

git clone  https://github.com/raseenaanwar/YT-CHAPTER-CREATOR  cd yt-chapter-creator npm install
add VITE_GEMINI_API_KEY to .env
npm run dev Free Gemini API key at aistudio.google.com.
This started as a way to save a few hours a week and turned into something genuinely useful. The gap between "AI gives you text" and "AI fits into a workflow" is bigger than it looks — and that gap is exactly what this fills. Happy to answer anything about the IFrame API integration, Gemini prompt engineering, or the Amplify setup.
Any opinions in this article are those of the individual author and may not reflect the opinions of AWS.
Enjoyed reading this content? Let the author know!

Your likes, comments, shares, and saves help creators reach more builders.

Loading recommendations

Loading article