Spaces:

fffiloni

/

CLIP-Interrogator-2

like 1.24k

Running on T4

If an image has text in it, the concepts in the text tend to leak into the caption. Here "eyes closed, the metaverse, asleep" don't really describe the image. (used a Persona 5 screenshot crop)

Input img:

#### Caption: a person sitting on a bed in a room, a screenshot, inspired by Munakata Shikō, vanitas, eyes closed, dialog text, the metaverse, asleep