Skip to content
Motif
Esc
↑↓navigate↵open⌘Jpreview
On this page

Ask about a picture

Caption, count, find or ask.

Synopsis

motif ask [question] [image] [options]

Example

A question about the picture, answered from the picture.

motif ask "how many vessels are on the shelf?" shelf.jpg
Three bottles, one bowl.
Three amber glass bottles and a white bowl on a stone shelf
The image

Options

Flag What it does
--caption Caption the image.
--detect <thing> Return boxes around this thing.
--point <thing> Return a point on every instance of this thing.
--read Transcribe the text in the image.
--safe Say whether the image is safe for work.
--tier <tier> Trade cost against quality: fast, balanced, quality.
--dry-run Validate and price the call without running it.

It also takes the common options.

Models and price

Motif picks the Model for the Tier. Prices are per image where a Model has one; any other shows how it is billed. A rate that depends on the size of the file or the length of the clip is a rate, not a total, so check it with a dry run.

Model Chosen at Price
Moondream 3 Query Balanced $0.40/M input tokens
Moondream 3 Caption Balanced, with --caption $0.40/M input tokens
Moondream 3 Detect Balanced, with --detect $0.40/M input tokens
Moondream 3 Point Balanced, with --point $0.40/M input tokens
GOT-OCR 2.0 Balanced, with --read $0.05/image
NSFW Checker Balanced, with --safe $0.001/image

Dry run

Check the cost before you spend. The output path is where the file would go.

motif ask "how many vessels are on the shelf?" shelf.jpg --dry-run --fields cost,costBasis,output
{
  "cost": 0.0020448,
  "costBasis": "projected",
  "output": null
}