How I Built an AI Test Case Generator with Multimodal Inputs

Why multimodal AI testing matters for QA Engineers

Most requirement artifacts are not clean Markdown. They arrive as PDFs, Word specs, screenshots, and short videos. As a QA Engineer and SDET, I needed AI Testing that starts from those real inputs — not from a demo prompt box.

This article explains how I shaped AI Test Case Generation into an installable product with AI with Human Review — a three-persona review loop (Test · Dev · Product Manager) — instead of a one-off LLM notebook.

The product pipeline

The generator follows a fixed chain:

• PDF / Word / TXT / Image / Video intake
• AI case generation
• Three-persona review loop (Test · Dev · Product Manager)
• Excel / Markdown / XMind export

Shipping as installable AI Test Automation

AI Testing only counts when teams can install it. Official run modes: Docker, local source, npm global install, and OpenClaw plugin.

Read the full flagship case study for architecture, My Role, metrics, and Try it now commands.