US 12,462,441 B2
Iterative image generation from text
Celeste M.B. Bean, San Francisco, CA (US)
Assigned to Sony Interactive Entertainment Inc., Tokyo (JP)
Filed by Sony Interactive Entertainment Inc., Tokyo (JP)
Filed on Mar. 20, 2023, as Appl. No. 18/186,752.
Prior Publication US 2024/0320867 A1, Sep. 26, 2024
Int. Cl. G06T 11/60 (2006.01); G06F 16/532 (2019.01); G06N 3/08 (2023.01); G06T 11/00 (2006.01); G06F 40/56 (2020.01)
CPC G06T 11/00 (2013.01) [G06F 16/532 (2019.01); G06N 3/08 (2013.01); G06T 11/60 (2013.01); G06F 40/56 (2020.01); G06T 2207/20092 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A method of automated text-to-image generation, the method comprising:
receiving a first image generated by an automated text-to-image generator from an initial text prompt, wherein the initial text prompt comprises a first set of words;
receiving, from an image-to-text generator in response to providing the first image to the image-to-text generator, a text description of the first image, wherein the text description describes visual content of the first image and comprises a subset of words that are not in the first set of words;
selecting, for inclusion in a second text prompt, a second set of words from (i) the first set of words included in the initial text prompt and (ii) the subset of words included in the text description of the first image;
modifying the second text prompt in response to user input;
submitting the modified second text prompt to the automated text-to-image generator; and
receiving a second image generated by the automated text-to-image generator from the modified second text prompt.