Visual input, with AI assistance
For learners who don't know the pronunciation yet, or any product without a pinyin keyboard: pick a stroke, narrow down by component and position, get the character. The widget on the right is calling the same public API your platform can call — this isn't a mockup.
Enter your API key (or start a free trial), then tap a stroke.
What the AI edition adds
Same engine and data as the stable version. Features below are switched on one by one; the stable page at visual-input.html is not affected.
① Learning from real use ● on
Each completed lookup (strokes tapped, parts chosen, the character picked, how long it took) is recorded anonymously — the raw material for the features below. No full text is stored.
② Part names & hints ● on
Hover or long-press a part to see its reading and a one-line memory hint (朿 cì vs 束 shù).
③ Stroke tolerance ○ planned
Tapped 丶 but meant 乀? When nothing matches, the engine suggests the stroke you probably meant, learned from real confusions.
④ Smarter candidate order ○ planned
Candidates ordered by what learners at your level actually pick, and by the text you are reading when a platform passes it in.
Call it directly
Same Bearer-key auth as the structural decomposition API, with an independent quota — it won't eat into your decompose usage.
curl https://api.chinesekey.org/v1/visual-input/components?strokes=h \ -H "Authorization: Bearer pk_live_..." curl https://api.chinesekey.org/v1/visual-input/positions?part=阝 \ -H "Authorization: Bearer pk_live_..." curl https://api.chinesekey.org/v1/visual-input/candidates?parts=阝@A02 \ -H "Authorization: Bearer pk_live_..."GET /v1/visual-input — full endpoint list & limits →