See · inside a photo · updated 22 October 2026

Spot

Find, count, blur or cut out things in a photo, or ask where to click in a screenshot. The answer is drawn on your picture.

See the recipes Models for /spot
  • 5credits a photo
  • 6tasks
  • 20 MBlargest photo
  • 4ways to redraw a result, free

Entry I

Start here

Run /spot with a prompt (what to look for, up to 500 characters), the image, and a task. Or right-click any message with a photo and choose Apps › Spot this: same panel, no re-upload. The models are Perceptron Mk1.5 and Perceptron Mk1.

Photos up to 20 MB are scaled so the longest side is 1568 pixels; bigger only costs time.

Entry II

The tasks

Find it

Boxes and numbers are drawn on your photo. What to find, e.g. every pigeon.

Count

Each one gets a numbered dot, with the total in the corner. What to count, e.g. people in the queue.

Blur or hide

Matches are pixelated. You can switch to boxes after. What to hide, e.g. faces, license plates, names.

Cutout

Each match becomes a transparent PNG, with an outlined preview. What to cut out, e.g. the dog.

Where to click

For screenshots. An arrow points at the control. What you want to do, e.g. turn on dark mode.

Ask about it

Answers in words and draws nothing. A question about the photo.

Entry III

The result

Marks are numbered with a matching legend. When the result has boxes or outlines, buttons redraw it another way with no new charge: Marks, Blur them, Crop sheet and Cutouts.

Entry IV

Asking well

Name the thing plainly

"Every pigeon", "the red car", "people in the queue". One kind of thing per run.

Say "every" when you want all

Find and Count look for every match; a description that fits one item finds one.

Screenshots for Where to click

Describe what you want to do, not the button: "turn on dark mode".

It says when nothing matches

The model is told not to invent a mark. An empty result means it did not see one.

Entry V

Every option

What you can type after /spot, read from the command itself. You can run it again after 10 seconds.

OptionWhat it doesTakes
prompt needed What to find textup to 500 characters
image needed The photo file
task What to do (default: find it) textFind it — box every match, Count — point to each one, Blur or hide — pixelate matches, Cutout — transparent PNG per item, Where to click — screenshot pointer, Ask — answer in words
model Perceptron model (default: Mk1.5) texttype to search
hidden Only you can see this (default: false — post in channel) true / false

Entry VI

Recipes

Requests you can copy into Discord and change to fit.

Count a crowd

Copy this

/spot prompt: people waiting in the queue image: (photo) task: Count

Why it works. A numbered dot on each person, with the total.

Hide faces before posting

Copy this

/spot prompt: faces and licence plates image: (street photo) task: Blur or hide

Why it works. Matches are pixelated; switch to boxes to check what was found.

Cut out a pet

Copy this

/spot prompt: the dog image: (photo) task: Cutout

Why it works. A transparent PNG of each match.

Entry VII

Tips

  1. I

    Crop to the area you care about before uploading; small things in a huge photo are harder to find.

  2. II

    Use the free redraw buttons instead of running it again.

  3. III

    Ask about it answers in words only; use it for "what is written on the sign?".

Entry VIII

Limits

  • Photos only, up to 20 MB.

  • Points (Count, Where to click) cannot be cropped, cut out or blurred.

  • 10 seconds between /spot commands.

Entry IX

Questions

Is the cutout exact?

It follows an outline the model draws, close to the edge but not pixel-perfect. Fine for stickers and mock-ups.

Does it work on video?

No, /spot is photos only.

Why did it miss some?

Small, hidden or blurry items are easy to miss. Crop closer, or describe them more plainly.

/imageImage /digestDigest

Where the advice comes from

Prompting advice is from these free public guides, matched to what ALMA lets you set. Prices and limits are read from the bot itself, as of 22 October 2026.

Try it

Type /spot in any server with ALMA, or in a direct message to the bot.

All guides

Something here wrong, or missing? Tell us.