Answer
Desktop web offers Upload Files for supported documents, images, videos and audio. Phone layouts offer choices such as Upload Photos and Upload Videos. Native People messages with AI off can use Photos / Videos.
If Text image analysis fails, Hashchats can automatically try one alternate reader using the same image and instructions. This also applies to scheduled Text requests. No provider selection is needed. Both readers may still fail or decline. Analysis uses ordinary Text tokens, without an Image-generation fee; audio and video retain their existing readers.
Image and Video generation use still-photo references, not uploaded videos. A URL in a media-generation prompt does not substitute for attaching its contents. Choose Text when you want an uploaded video analyzed.