BoardSnap handwriting recognition: what to expect.
Short answer
BoardSnap uses Apple VisionKit for on-device text detection combined with a multimodal AI model to read and interpret whiteboard handwriting. In good lighting with clear, printed writing, accuracy is very high. Heavily stylized cursive, tiny text, or poor lighting will degrade results — but BoardSnap AI reads context, not just characters, so it can often infer meaning even when individual letters are ambiguous.
## Two layers of reading
BoardSnap doesn't rely on a single OCR pass. The pipeline runs in two stages:
Layer 1 — Apple VisionKit. After the perspective correction, VisionKit runs on-device text detection. It identifies text regions on the board and does an initial character-level read. VisionKit is trained specifically on Apple's hardware and has been improving with every iOS release.
Layer 2 — Multimodal AI. The processed image goes to BoardSnap AI (running on Anthropic Claude or OpenAI's vision models). Unlike traditional OCR, these models read the image with context — they understand that a box with an arrow pointing to another box probably represents a dependency, that a circled word is probably important, that a list of bullet points under a heading is probably a set of tasks. This contextual reading is what makes BoardSnap produce action items instead of just a wall of recognized text.
## What affects accuracy
Improves accuracy:
- Good ambient or natural lighting on the board
- Dark marker on a clean white surface
- Printed letters (not cursive)
- Large enough text (thumb-sized letters or bigger)
- Standing directly in front of the board (VisionKit corrects perspective, but extreme angles reduce clarity)
Degrades accuracy:
- Low-light rooms — flash helps, but shadows can still cause issues
- Worn or pale whiteboard markers
- Heavy cursive or stylized fonts
- Very small text or dense tables with tiny cells
- Board glare from overhead lighting
## Can BoardSnap read cursive?
Yes, to a degree. Light cursive with clear letterforms works reasonably well. Heavy, looping personal handwriting is harder. If your team writes in block letters on the board, accuracy will be consistently high.
## What happens when a word is misread?
BoardSnap AI uses context to fill gaps — if it misreads a character but the surrounding words make the meaning clear, the summary will still be correct. You can also edit any action item inline after the analysis completes.
## Honest benchmark
BoardSnap is not a perfect OCR engine and doesn't claim to be. It's a thinking tool: even with minor transcription errors, the summary and action items are usually directionally correct. Think of it as a very smart reader who might misread one word in fifty but always gets the gist.
Frequently asked
Does BoardSnap work in low light?
It works, but accuracy drops. Use your iPhone's flash or turn on the lights. The cleaner the photo, the cleaner the output.
Can I correct mistakes in the output?
Yes. Every action item is editable inline after the analysis. Tap to edit, fix any misread text, and save.
See it work in ten seconds.
BoardSnap is free on the App Store. Snap a board — get a summary and action plan.