AI is not nearly powerful enough for my image-describing needs; CW: long (1,844 characters), Fediverse meta
Artikel ansehen
Zusammenfassung ansehen
As long as AI can't always and 100% reliably...
- tell from looking at a picture where exactly it was made, even if that place is in an obscure 3-D virtual world
- look at any item in a 3-D virtual world, identify it correctly, no matter how small it is, tell who has made it and judge whether it's important to mention who has made it or even explain who that person was/is
- transcribe text that's less than one pixel high
- transcribe text that's partly or even mostly obscured and even transcribe the invisible parts of the text
- transcribe text that's only partly within the picture and then correctly include the out-of-frame parts
- look at an in-world sales box, rattle down the contents and, if necessary, explain them
- describe images within images or even images within images within images, no matter how tiny, at the same level of detail as the image itself
- not only identify a teleporter as such, no matter what it looks like, but also rattle down a full list of chooseable destinations and explain what these destinations are
- look at an in-world door and tell if it's scripted and how (door-opening script vs teleport script to the other side of the door vs teleport script to entirely elsewhere)
- tell an avatar from an NPC from an animesh figure from a static figure
- identify avatars, no matter how small they are in the picture, and judge whether it's important to mention who they are or even go into more detail
- tell from looking at a picture from an event which event is happening there, no matter how obscure and niche, and, if the event organisers are in the picture, identify them
...I'm not interested in using it for image descriptions. Thanks, but no, thanks. I'll go on writing them myself.
#
ImageDescription #
ImageDescriptions #
AltText #
AI #
Long #
LongPost #
CWLong #
CWLongPost
