Scaling Educational Content With Human-Led AI

How Blended Teaching and Between Pixels Built an AI-Powered Production System That Cut Editing Time by More Than Two-Thirds

The Ask

Blended Teaching wanted to produce more educational content, but editing and graphic creation had become the primary bottlenecks. Each chapter required approximately 40 hours of production time, including roughly 30 hours of editing, limiting turnaround speed and available studio capacity.
 
Between Pixels was uniquely positioned to solve the problem because we had proprietary knowledge of how these educational edits are structured, along with a library of successful human-edited chapters that demonstrated the creative and instructional decisions behind them. We used that expertise and historical editing data to train AI models to recognize those same patterns, helping reduce the total production budget to 20 hours per chapter without sacrificing quality, accuracy, or the human judgment required to create an effective learning experience.

Video
Production

Higher Edu

The Results

To address the production bottlenecks, we developed a custom AI-driven rough-cut tool that successfully reduced editing time from approximately 30 hours to fewer than 10 hours per chapter, a reduction of more than two-thirds. This allowed Blended Teaching to move beyond exploring AI potential and begin using a functional production system that measurably improves studio capacity. Additionally, we are building a cloud-based graphics engine with full AI integration; this creates a connected workflow that streamlines chapter production while maintaining high consistency across the content library.

The Work

Between Pixels partnered with Blended Teaching to develop two connected, AI-powered production solutions.

First, we created a custom rough-cut tool that automates much of the initial video assembly process. The tool organizes footage, builds the chapter structure, and produces a working edit that gives the human editor a significantly more advanced starting point.

At the same time, we began developing a cloud-based graphic generation engine with full AI integration. The system analyzes lecture transcripts, identifies where visuals can improve comprehension, and generates graphics based on the instructional and editorial decisions established by experienced human editors.

Together, these tools automate repetitive production tasks while keeping people responsible for instructional strategy, accuracy, creative direction, brand alignment, and final polish.

The Future of Scalable Educational Content

The long-term value of the system extends beyond Blended Teaching’s professionally produced chapters. The cloud-based graphics engine is being designed to eventually connect directly with the user-generated content side of the Blended Teaching platform.

An educator will be able to upload a simple screen recording of themselves teaching from a basic slide deck — or even provide voice-only narration. The platform can then use approved, AI-generated footage of the lecturer to create a complete on-camera presentation, enhance the instructional content, and add custom graphics that support the lesson.

What begins as a straightforward screen recording can be transformed into a polished, professionally produced video without requiring cameras, a studio, an editing team, or an advanced production workflow.
 

The result is not AI replacing the people who understand the material. It is a scalable system that preserves the educator’s voice and expertise while giving them the tools to reach more students, create more content, and produce it in significantly less time.

Is the work made by AI or by people?

It depends on the project, but usually both. In some cases, our team creates and distributes all of the original content. In other cases, we design the system, templates, and kit of parts for AI to produce at scale. In some cases, we need to use AI to create visuals that don’t exist.

Yes. Our team leads the production of original content, and our AI platform (SEMBL) is trained in your brand, customers, business, and industry.

It varies by project, but large projects can see a reduction in production costs of 10x or more.

We use enterprise models under no-training policies; your brand data isn’t used to train shared models.

Scroll to Top