OpenAI just released ChatGPT Image 1.5, and while it is better, it’s not as good as Google’s Nano Banana Pro.
With some initial testing, I found that:
🔷 ChatGPT isn’t good at adapting my facial features to different scenarios. Sometimes, the guy in the photo looks completely different from me. Google has a lot of deep learning data from Google Photos so it’s much better at generating deepfakes (or Pointless LinkedIn Selfies).
🔷 ChatGPT is still slow. It takes over a minute to generate an image.
🔷 There’s no way to increase the resolution of the ChatGPT Image, even in OpenAI’s Playground. Google lets you create images up to 4K resolution in AI Studio. A ChatGPT square image is still only 1024 x 1024 pixels, but Gemini provides double the resolution..
🔷 Still no animation capability. C’mon, Google already offers free text-to-video animation in Google Whisk.
Anyway, here is a comparison where ChatGPT didn’t mess up my face. As you can see, Google follows the prompt better, especially with regards to the cats. I wanted cats in suits, not robot cats.
My prompt (with uploaded profile photo): “Put me in a 1950s style movie painting. I am in a yellow spacesuit, holding a laser gun, running away from cats in mecha exoskeletons, background is a city. The title of the movie poster is “Run, Ian, Run!”. Vertical aspect ratio. Vibrant colors. Dramatic camera angle.”