Upload quiz photos, enter your answer key — get scores, letter grades, and class analytics instantly. On-device AI, no data leaves your browser.
One photo per student. Clear lighting, flat paper, answers circled or written in large print works best. JPG, PNG, WebP, HEIC — any orientation.
Correct answers in order, one per question. Case-insensitive. Supports A/B/C/D, True/False, numbers, or short words.
Or paste all answers (one per line):
Paste auto-sets the question count to match the number of lines.
Customize how scores and letter grades are calculated.
This tool uses SmolVLM-256M-Instruct, a compact vision-language model from HuggingFace, running entirely inside your browser via Transformers.js with WebGPU or WASM acceleration. Your images are never sent to any server.
Tip for best accuracy: flat paper, good lighting, and clearly written/circled single-character answers (A/B/C/D) yield the highest read rate. Any answer the model marks as ? should be verified manually — re-run individual sheets by removing and re-uploading after improving the photo.
? in the table, check that question manually. By default, ? answers are treated as skipped (not wrong) — check "Count unread answers as wrong" to change this.