Case study · AI production

12,000 video solutions, made by a pipeline instead of a studio

Recording a video for every past-paper question would take a studio years. We built a pipeline that turns each worked solution into a narrated video lesson, checks it, and publishes it to the student portal, at a cost per video a studio could not come near.

The client

ExamYard (examyard.com), a Cambridge exam-prep publisher. ExamYard is built and run by the Inertia Automate team. It is our own work, so we can show it in full.

What needed to change

More than 12,000 past-paper questions, each needing a clear video explanation.
Lessons that show science in 3D rather than flat diagrams.
A cost per video low enough to cover every question, not a sample.

What we built

No.PartWhat it means in practice

01Script to videoEach worked solution turned into a narrated, step-by-step video lesson with on-screen working.

02Bulk runs on GPUsHeavy work run in bulk on rented GPU servers, then shut down, so the cost is paid only while videos are being made.

033D models in lessonsInteractive 3D models of hearts, cells, circuits and more, placed in over 41,000 lessons.

04Checks before publishingAutomatic checks catch missing or broken videos before students see them.

05PublishingVideos uploaded and linked to their questions in the student portal, with checksums to confirm each file arrived intact.

How it runs now

12,117 of 12,286 solution videos were live by 2 November 2026, in the ExamYard Student Portal.

Built with: Python, Text-to-speech, Video rendering, GPU cloud (RunPod), 3D models (glTF), PHP / MySQL portal.

More case studies

Free tools

Free SEO checkerSee how Google, ChatGPT and Copilot read your website. 20 checks, a score and a fix for each, in 20 seconds.Check my website →llms.txt generatorDraft the file AI assistants read to understand your site, from your own sitemap, in under a minute.Make my llms.txt →

Book a meeting

Free 30-minute call