Second Look

Point your camera at something. One model explains what is in the photo and pins each claim to a spot. A second model, shown only that spot, says whether it is really there.

Made with pollinations.ai

Your photo

Sign in with Pollinations to send a photo. You pay with your own Pollen; the default models are free-tier.

Recorded photos: try it without signing in

How it works

Explained and pinned

The photo goes to z-ai/glm-5.3-flash with a lettered grid drawn on it. It names 3 to 5 things it sees, each pinned to one grid cell, with how to spot it and one fact the photo cannot show.

Checked blind

Code crops each pinned cell. openai/gpt-6-luna, from another model family, sees only that crop, says what it sees, then answers yes or no: is the claimed thing there?

The checker is tested too

Every photo also gets a decoy question about something that is not there. If the checker says yes to the decoy, its confirmations on that photo are marked weak.

Green: confirmed. Red: disputed, the second look did not see it there. Grey: no clear answer. "Did you know" facts come from the model's knowledge and are not checked.