Your site needs alt text for thousands of images. What should you use?
Why D is the answer
A multimodal model that generates descriptions
A multimodal model can look at an image and generate alt text or an extended description for it.
In detail
- A multimodal model can describe each image.
- Review for accuracy on important pages.
Use case
An illustrative scenario, not a real incident.
A site has thousands of images with no alt text. A multimodal model drafts a description for each, and editors review a sample.
See it in different ways
flowchart TD
A{"What do you need?"} -->|"New image or video"| B["Generation model"]
A -->|"Change part of an image"| C["Inpainting with a mask"]
A -->|"Describe an image"| D["Multimodal model"]
A -->|"Structured output from video"| E["Content Understanding"]flowchart LR S0["Image"] --> S1["Multimodal model"] --> S2["Draft alt text"] --> S3["Review"]
Key point: Multimodal models can describe images.
Why the other options are wrong
- A. A keyword search across image file names
- File names rarely describe an image.
- B. An embedding model on its own
- Embeddings are vectors, not text.
- C. A speech synthesis voice that reads each file name aloud to reviewers
- Reading file names aloud does not describe images.