Every image and video below came straight out of the API, with the exact prompt beside it. Hover a video to play it, click anything to see the prompt and try it in the playground. See one character across ten scenes, and before and after examples.
One character, made once: a generated portrait saved with POST /v1/characters, then passed as character to ten image calls and one video. Each prompt sets its own pose, expression, camera angle, lens and light (hover a picture for its prompt); the face, hair and the mole under her eye carry over, with no retouching. Anything a prompt leaves out, like the top in two of these, comes from the portrait. Images $0.09 each on spicy-image-1-pro with a character; the clip is 5 seconds at 720p on spicy-character-video-1 ($1.00).











spicy-image-reference-1 describes the picture in words, leaving out identifying facial features, and draws a brand new image from that description. The picture itself is never edited, so the result has its look but not its person. $0.12.


spicy-upscale-1 at 2x took the pool picture from 1024x1536 to 2048x3072. The close-up shows the same eye from both, at the same size. $0.02, or $0.04 at 4x.

Left: original. Right: 2x.
mask_url on POST /v1/images/edits, prompt "Change her red knit scarf to a cream cable-knit wool scarf, keep everything else the same". White in the mask changes; everything under black stays the original, pixel for pixel. $0.09.


