Testing the new gpt-image-1 model
Available Settings
- Quality: auto, high, medium, low
- Size: auto, 1024x1024, 1024x1536, 1536x1024
- Format: jpeg, png, webp
- Moderation: auto, low
- Background: auto, transparent
Image Generating
Our prompt:
A young girl, in yellow dress, looking at a notice board.
The notice board says: We are closed on Monday 28/4/2025
In studio ghibli style
1) 1024x1024, high setting
2) 102x1024, medium setting
Image Editing
1) Change the facial expression
Make the little girl look a bit disappointed on her face
2) Multiple instructions.
Translate the text on the notice board to simplified Chinese.
Then replace the original text with the translated text.
Items extraction
1) Extract items from a photo taken by the camera. Using got-image-1 high setting.
Extract the pastries in the photo.
Design a menu-like image based on these pastries.
Source image:
Output Image:
The image shows one extra item, but the other items are correctly extracted.
2) We try to add a price to each of the items.
Please add a price tag to each item in the image.
All of them are RM 3.50, except the first item, which is RM 4.80
gpt-4o-2024-08-06
2025-04-26 19:24:54
The blog post examines the new gpt-image-1 model, highlighting its settings for quality, size, format, moderation, and background. It showcases image generation, editing capabilities, and item extraction functionalities, such as changing facial expressions and translating text on images, as well as extracting items from photos for menu creation.
Chrome On-device AI
2025-05-27 23:45:30