More than text
An AI app that answers a question about an uploaded photo uses both image and text. Another might take a spoken request and reply aloud, or create an image from a written description. These are all examples of multimodal AI.
It can still miss things
AI can miss small details in an image or words spoken over background noise. A confident answer can still be wrong. Apps should test each type of content, protect private photos and recordings, and offer accessible alternatives.