Describe image

Read text out of an image, or write alt text and captions for it.

POST /api/v1/describe-image MCP describe_image aiAI

Vision models are the priciest option here — roughly 100+ neurons per image.

Playground

Base64-encoded image bytes (data: URLs accepted).
Image URL (PNG, JPEG, WebP or GIF).
What to produce.
Advanced options (2)
Response length cap.
Custom instruction. Overrides task.
Runs against this deployment's live API.

Result

Fill in the form and hit Run.

Call it from code

How it works

A vision model with four presets:

  • ocr — transcribe every piece of text, preserving reading order
  • describe — a detailed description of the scene
  • alt-text — one accessibility-grade sentence
  • caption — a short editorial caption

Pair it with screenshot to let an agent "look at" a page it cannot parse, or with task: "ocr" to pull text out of scans and diagrams.

Parameters

ParameterTypeDefaultDescription
url string Image URL (PNG, JPEG, WebP or GIF).
imageBase64 string Base64-encoded image bytes (data: URLs accepted).
task `describe` | `ocr` | `alt-text` | `caption` "describe" What to produce.
prompt string Custom instruction. Overrides task.
maxTokens integer 512 Response length cap.

Response

The model's text output.

ParameterTypeDefaultDescription
text string Result.
task string Task that was run.
model string Model used.

Pairs well with