Answer
- Image and Video use images in the replied message as references. If it has no images, its text can be added to your prompt. Video uses the first reference image.
- Image and Video do not analyze the reply’s audio or video as general chat context.
- Audio can incorporate replied-to text, but does not accept its media as generation input.
For precise control, attach the reference directly and check the composer preview before sending. These rules differ from Text AI’s richer selected-message context.