Moondream Vision — Playground

Drop images from your computer, ask a question, and Moondream will answer. Tip: for people detection, prefer wording like "Is a human visible?" — some short prompts can return empty. How-to (face workflow + API): /instructions. For the raw API, see Scalar, Swagger, or openapi.json.

Reference vs up to 10 candidates

What this does: pick one reference photo, then 1–10 candidate photos. Each candidate is sent to the model together with the reference; you get a yes / no / unclear label per row plus the model text (same as POST /v1/face/compare/upload). This is a soft visual check, not a certified biometric matcher — read /instructions for the disclaimer.

Reference (1 image)

Reference — click or drop

Candidates (1–10 images)

Candidates — click or drop (multiple)

Max tokens / compare

Temperature

Images

Click to choose files or drag & drop here

PNG, JPEG or WebP · up to 16 images · 20 MB each

Prompt

Max tokens

Temperature

top_p

top_k

Thinking mode

Response

(nothing yet)

Raw JSON