visual-qa

Automate Godot visual game testing by validating screenshots and frame sequences against reference images and acceptance criteria. It detects visual defects, motion anomalies, and UI cropping issues using API-based visual question answering.

Detect jitter, teleportation, stuck entities, and camera timing mismatches by reviewing consecutive Godot frame captures. Dynamic motion verification analyzes 2 FPS frame sequences to identify animation anomalies blocking acceptance.

Check Godot UI screenshots for cropping without a reference image using question-based visual analysis. You provide targeted questions about the UI crop location, and the VQA backend evaluates the screenshot to answer acceptance-blocking queries.

Visual question answering for Godot requires Python dependencies including Pillow for image processing, alongside google-genai and openai libraries. These enable the configured VQA backend selected via the .godotmaker/config.yaml file to process screenshots.

Verify Godot scene acceptance criteria visually by comparing a single game screenshot against a per-scene reference image. You provide Goal, Requirements, and Verify context to the skill, which routes execution through native inspection or a Python API.