Point the camera. Get a micro-lesson.
A pupil points at a flower, a machine part or a book. The AI recognises it and builds a 3-question micro-lesson in 10 seconds — tuned to the pupil's VARK style and year group.
What's new
GPT-4o Vision recognises
Object detection on device + vision model behind — works on objects, animals, plants, text and diagrams.
3 questions, one level up
The AI calibrates one step above the pupil's current mastery — stretching without frustration.
VARK variants automatically
Visual = image + diagram, Auditory = TTS, Read/Write = transcript, Kinaesthetic = 'find three examples in the classroom'.
10 seconds from click to quiz
Optimised with gpt-4o vision + gpt-4o-mini for Q-gen — cheap enough for daily use.
PII-scrub on every prompt
Images stored in a private storage, prompts run through aiSafeFetch. No faces, no names.
Counts in My learning
Every scan lands in the pupil's knowledge map and awards XP. Teachers see what was scanned and how the pupil answered.
Examples from the classroom
Science, Year 4
Pupils walk the playground and scan plants. Each scan gives a mini-quiz on photosynthesis.
Design & technology, Year 7
Scan a tool in the workshop — get safety rules and a quick check before use.
English, Year 5
Scan everyday objects from home and get vocabulary quiz + pronunciation coach in English.
Food studies, Year 6
Scan a fruit or vegetable — get nutrition, season info and a mini-recipe.
Security & compliance
Camera on demand only
Camera activates only when the pupil taps 'Scan'. We don't store background feed, faces or location metadata.
Parental consent under 13
Camera use for pupils under 13 requires explicit parental consent (child_data_consents) before the AI can be called.
PII-scrub on every prompt
Image + text context runs through aiSafeFetch and is rejected (422) if CPR, names or emails are detected.
Our own OpenAI key
No Lovable AI Gateway. We call OpenAI directly with store:false and hashed user-id — no third-party training.
Frequently asked questions
Which objects can the AI recognise?
Almost anything: plants, animals, tools, food, buildings, text, diagrams, artworks. gpt-4o vision is trained on a broad domain — and when unsure, it says so openly.
Can pupils scan themselves or other pupils?
No. We reject images with detected faces via a pre-filter before the prompt is sent — and the AI never answers questions about people.
What does each scan cost?
Less than a penny. We combine gpt-4o vision with gpt-4o-mini for quiz generation, keeping marginal cost low enough to include in the base subscription.
Does it work offline?
The camera capture works offline, but the AI call needs a connection. Scans queue locally and sync when the device is back online.