Question 1

How do you build an AI course generator SaaS with Next.js 13 App Router?

Accepted Answer

I start with a fresh Next.js 13.4 project using the App Router, then build the core flow: create course → generate units/chapters with OpenAI → enrich chapters with YouTube videos and transcripts. I keep the UI fast with shadcn/ui + Tailwind, and I structure routes using the app directory (page.tsx per route). The key is treating it like a real product, not just a demo.

Question 2

How do you generate quiz questions from YouTube transcripts using OpenAI?

Accepted Answer

For each chapter, I search YouTube, grab a relevant transcript, and feed that transcript into OpenAI. From that input, I generate a summary and concept-check questions that are actually tied to what the video says. I run generation asynchronously per chapter, so each chapter can complete independently and you see progress as they turn green.

Question 3

How do you add Google login in Next.js 13 with NextAuth and Prisma?

Accepted Answer

I use NextAuth with the Prisma Adapter, then set up the standard User/Account/Session tables in schema.prisma. In my callbacks, I attach the database user id and credits onto the JWT, then map them into the session so the frontend can read them. You’ll also need to create a Google OAuth client ID/secret in the Google Cloud console.

Question 4

How do you implement a free credits system for an AI SaaS?

Accepted Answer

I add a credits integer on the User model and default it to 10, so every new user starts with 10 generations. Each time you generate a course, I decrement credits and display the remaining count in the UI. Once it hits zero, you can gate course creation and push users to upgrade.

Question 5

How do you integrate Stripe subscriptions for unlimited usage?

Accepted Answer

I use Stripe to create a subscription checkout, and after payment I rely on a webhook back into the Next.js app to update the user’s subscription state. In the demo, once the webhook hits, the UI reflects that you’re a pro user and the generation limit disappears. This is the exact pattern you can reuse to start charging for your SaaS.

Question 6

Why not deploy this AI SaaS on Vercel?

Accepted Answer

For long-running AI workflows, you can easily exceed serverless execution limits. I call out that Vercel can time out quickly on the free tier, and course generation can take longer than that because you’re hitting multiple AI endpoints. That’s why I show self-hosting on a VPS so your functions can run as long as you need.

Question 7

How do you self-host a Next.js app on DigitalOcean with a custom domain and SSL?

Accepted Answer

I rent a DigitalOcean VPS, deploy the Next.js app there, then point my domain to it and add SSL so you get HTTPS on your custom domain. The goal is to make it production-like: stable hosting, no serverless timeouts, and proper security. I also show setting up GitHub Actions so deployments are automated.

Build & Deploy: Full Stack AI Course Generator SaaS with NextJS 13, OpenAI, Stripe, TailwindCSS

🛍️ Products Mentioned (10)

GitHub Repo

54 prisma db.ts file

00 schema.prisma template code

45:00 lib/gpt.ts

58:42 lib/unsplash.ts

59:51 lib/youtube.ts

48:47 .github/deploys.deploy.yml

Strict JSON GPT repo

GitHub

Website

About This Video

Frequently Asked Questions

🎬 More from Elliott Chong