Inspiration
Learning online is scattered. You end up with ten tabs open — a YouTube video that's six years old, a blog post missing the one step you actually needed, a doc that assumes you already know half the material. And even the good tutorials share the same flaw: someone reads a script while a finished project sits on screen, already built. You watch, you nod, you feel like you understood it. Then you try to build it yourself and realize you never actually watched anything get built — you watched someone perform having already built it.
We wanted something different: describe what you want to learn and where you're starting from, and get back a real lesson — built for you, at your pace, watched as it comes together rather than handed to you finished. So we built LearnForge AI.
What it does
You tell LearnForge AI what you want to learn and your skill level. It generates a video lesson: a real explanation, narrated with AI-generated voice, code and content appearing on screen in sync with what's being said, captioned throughout.
Every generated video lands in a public library where you can browse what others have learned, and like, comment on, and bookmark the lessons that are useful to you. Instead of hunting across scattered tutorials, you get one place where the exact lesson you need already exists — or gets created the moment you ask for it.
How we built it
LearnForge AI is built on Next.js, TypeScript, and Tailwind CSS, with Firebase handling authentication, data storage, and the library's social features (likes, comments, bookmarks), and Google Cloud plus the Gemini API powering lesson generation. Gemini plans the explanation and content for each request and drives the generation pipeline; that output is composited into a narrated, captioned video synced to what's on screen. Deployed on Vercel.
Challenges we ran into
Getting the AI to generate something that teaches, not just explains, was the core problem. It's easy to get a model to produce a correct explanation — it's much harder to get narration, on-screen content, and captions to feel genuinely synced rather than just technically aligned. A video where the voice is half a beat off from what's on screen stops feeling like a lesson and starts feeling like a glitch.
We also had to figure out how to keep generated content structured and level-appropriate across wildly different requests, instead of producing something generic that reads the same regardless of what someone actually asked for.
Accomplishments that we're proud of
Type in what you want to learn, and minutes later there's a real narrated video of it — built for your level, captioned, ready to watch — that didn't exist anywhere until you asked for it. Getting that loop working end to end, and getting it to feel personal rather than generic, is what we're most proud of.
We're also proud of the library: LearnForge AI isn't a generator you use once and forget — it's a growing place where lessons stick around, get discovered, and get engaged with by other learners.
What we learned
We learned that generating a plausible-sounding explanation is easy — the hard part is generating something structured like a real lesson, paced the way a person would actually need it. We also learned how much "feel" matters in AI-generated video: sync between voice, visuals, and captions is the difference between something that feels like a mentor and something that feels like a glitch.
What's next for LearnForge AI
Remix: let anyone take another user's lesson and regenerate it into their own version, tuned to their level. (Move this back into "What it does" if it's already live — this is worth confirming since it's a strong differentiator.) Quizzes and assignments after each lesson, to turn watching into practice. Progress tracking, so a learner's history across lessons builds into something that feels like a path, not a series of one-offs. Voice tutoring and the ability to pause a video mid-lesson to ask a question. Multilingual support, certificates, and deeper personalization based on what someone's already learned.
Built With
- ai
- backblaze
- cloud
- gemini
- nextjs
- postgresql
- prism
- recall
- tailwindcss
- typescript
Log in or sign up for Devpost to join the conversation.