Describe Image
Describe Image sends one or more images to an AI model with vision capabilities and returns what the model sees in them. With the Prompt field you decide what you want to know: a general description, whether a specific product appears, the condition of an object or the text on a label.
If you connect several images at once, the model analyses them together and returns a single answer for all of them.
When to use it
- Write product descriptions from your catalogue photos.
- Review the photos a customer sends to assess damage or an incident.
- Check whether a photo meets your requirements (white background, visible logo…).
- Let an agent look at the images it receives in a conversation.
Configuration
| Field | What it does |
|---|---|
| Images | The image or images you want to analyse. Connect the file coming from another component. |
| Prompt | What you want the model to do with the images. Default: “Describe the image in detail.” |
| Model (required) | The AI model that analyses the images. It must be a model that can see images. |
Very large images are reduced before being sent to the model.
This component supports Batch Mode, to describe each image in a list separately, and it can also be used as an agent tool. See Key concepts.
Output
The Data output returns one result with:
- image_description: the model’s answer.
- images: the name and type of each analysed image.
- num_images: how many images were analysed.
- prompt: the instructions used.
Example
An insurer receives photos of damage to a vehicle.
- Connect the received photos to Images.
- In Prompt, write: “List the damaged parts of the vehicle and say whether the damage looks minor, moderate or severe.”
- Connect a vision-capable model to Model.
- Send the description to the claims handler together with the case file.
Good practices
- Be specific in the Prompt. If you need an answer in a fixed format (for example, a list), ask for it explicitly.
- If you want one answer per image, use Batch Mode instead of connecting them all at once.