Drawing Scope exists because he is a beginner artist who wanted the feedback it gives, and he carried the pitch for it alone. He also has a sharp eye for what a model leaves out, and knows what vagueness costs — a one-word prompt returns nothing usable, a specific one returns a picture. Building for a user you understand completely is the strongest position a designer can start from.
Drawing Scope
Nathan Yeh · Earmias Biazen
The problem they picked
Learning to draw the figure alone is mostly guesswork. A tutorial shows you a finished result; it cannot tell you that the shoulders sit too high or that the pose carries no weight. The specific, unflattering feedback that actually moves a beginner forward is the one thing the internet will not hand you for free — and Nathan knew exactly what was missing, because he was the beginner asking for it.
Drawing Scope closes that loop: a reference to work from, a canvas to draw on, and a critique that scores gesture, construction, proportion, and anatomy separately — one strength and one fix, every time.
“As a beginner artist myself, I often see lacking opportunities to improve online. How do I improve my structure? How can I improve the way I invoke the gesture in my drawings?”
Drawing Scope, at the showcase
How it was built
Computer Vision Basics
A vision model reads the redraw against its reference and scores gesture, construction, proportion, and anatomy separately.
Prompt Engineering
A 70-line critique prompt that has to return a score out of 100 per skill, plus one strength and one thing to fix.
Front-End Development
The redraw surface is the browser's own Canvas API, with practice history kept in local storage.
Canvas APIBrowser API
n8nAutomation Platform
GPT-4o visionAI Vision Model
Also in the buildPer-skill critique scoring · Plan–draw–review practice loop · Bundled 16-photo reference library
