Case 01 · B2B SaaS

IntelligentVideos.ai: three AI products, one account and a single credit system

IntelligentVideos.ai is a B2B SaaS that turns URLs and documents into AI avatar videos, GEO packages and publish-ready pages. I build and maintain its web product and backend: Clerk authentication, credit billing with Stripe, generation queues with BullMQ and Redis, and Claude integrations, on Next.js, Express and PostgreSQL.

Updated:

IntelligentVideos.ai
Live product Visit IntelligentVideos.ai ↗

What is IntelligentVideos.ai?

It is a platform for businesses that care about how their content performs in AI search. It brings three products together under one account:

Video Creator

Upload up to 5 PDFs and 5 URLs, or describe a topic. The platform writes three scripts for you to pick from or edit, then generates the video with an AI avatar.

GEO Optimizer

Paste a URL and get an assessment of the page plus the code and copy ready to paste, not just a report.

Page Transformer

Paste a URL and get a mockup of your page rewritten, a new page on a topic your competitors cover, and a landing page for the video.

All three spend from the same credit balance, which comes included in the subscription plan.

What problem did it solve?

A product like this is not one model call and one response. According to the product’s own site, a video and its deliverables take minutes to be ready, and several external services take part along the way: AI models, avatar generation and video hosting.

On top of that come three products drawing from the same balance, subscriptions with recurring payments, and services billed by usage. The challenge was making all of it reliable and measurable: a long generation should not be lost when something fails, every credit should be deducted correctly, and the cost of running it should be visible.

What did I build?

I build and maintain the web product and the backend. That work includes:

  • The backend design, with Clean Architecture.
  • An event-driven job pipeline with BullMQ and Redis for long-running generation workflows, with retries.
  • Authentication with Clerk.
  • Credit billing and subscriptions with Stripe.
  • Integrations with Claude and Amazon Bedrock, the backend’s two AI providers.
  • Integrations with Tavus for the avatar and Bunny Stream for the video.
  • Cost observability, so usage is measurable.
  • The Next.js web product where the three products live side by side.

How is it put together?

The core idea is that generation does not happen inside the web request: it goes to a job queue, and the result arrives when it is ready.

  1. Web

    Next.js 16. You sign in (Clerk), pick a product, review the scripts and receive the deliverables.

  2. API

    Express 5 with Clean Architecture: the product rules (accounts, credits, jobs) stay separate from the external services.

  3. Job queue

    BullMQ on Redis. Each generation is a long-running job that moves forward through events and is retried when a step fails.

  4. External services

    Claude and Amazon Bedrock for AI, Tavus for the avatar, Bunny Stream for the video and Stripe for payments.

  5. Data and cost

    PostgreSQL as the database, plus cost observability so usage is measurable.

Which decisions mattered?

Queues instead of waiting for a response

A generation that takes minutes should not depend on an open request. That is why it is asynchronous: it enters a job queue on BullMQ and Redis, and when a step fails it is retried instead of losing the whole job.

One credit balance for three products

Video Creator, GEO Optimizer and Page Transformer share one account and one credit system. For the customer it is a single balance; for the backend, one way to charge, deduct and measure usage instead of three.

Authentication and payments handled by specialists

Sign-in runs on Clerk and payments on Stripe. As the product’s own site states, the platform does not store passwords or card numbers.

Two AI providers in the same architecture

The backend works with Claude and with Amazon Bedrock. Clean Architecture keeps those integrations at the edge of the system, apart from the product rules.

Cost is observed too

Every generation uses external services billed by usage. Cost observability makes that usage measurable, not a guess.

What does it solve today?

Video Creator, GEO Optimizer and Page Transformer share one account, one credit system and an asynchronous flow that delivers video, code and content.

You can see the live product on its public site: a free GEO scan with no signup, 12 avatars for the video, and deliverables in five languages (English, German, Spanish, French and Portuguese).

What is it built with?

Each link goes to the official documentation.

Have a similar project?

If your product needs AI generation that takes minutes, credit-based billing or integrations with several providers, message me and tell me where it stands. If it fits what I do, I will tell you how I would approach it.

Message me on WhatsApp See services and pricing

Another case

Case 02 · Health marketplace

Recovery Care Solutions

Recovery Care Solutions: a React Native app and a new backend, route by route

Read the case →

See all projects