This is an AI image generation comparison for
image-to-image 
prompt:
A bigfoot wearing lime-green sunglasses holds this soda can in one hand as if advetising it, behind is a blurred out tropical beach...
Log in to see full prompt.
Tested: June 27, 2025
Official demo used on HuggingFace
Tested: July 17, 2025
If you provide a product image of the soda can and do not mention that it should be small, model draws a soda can the size of bigfoot. 'Photorealistic, cinematic shot' did not influence the style.
Tested: August 18, 2025
Nicely done from the 1st try, even transferred the watermark url which was on the original soda can image as a bonus)
Tested: August 27, 2025
This image editing test asks the model to create a new image which will contain a product form our input image. In this case, a soda can with our text. So it tests both the reference to image abilities and text handling.
Check out the results from FLUX.1 Kontext (FLUX Kontext [Dev]) vs Replicate (SeedEdit 3.0) vs Qwen Image & Video Generator (Qwen Image Edit) vs Freepik (Gemini 2.5 Flash) for similar or identical prompts side-by-side.
