April 21, I created stickers of Cinderella in a cartoon style. Despite my effort, I still had to scale up the image since the bots did not perform as expected. It seems that machines still struggle to accurately describe the complexity of human beings visually.
I think the bots are only smart if we provide good instruction in the prompts. We have to find out how well the bots can understand us to carry out the tasks precisely.
I encountered some underperformance in AI work while using Midjourney. Essentially, the initial image quality was subpar until I upscaled it. This upscaling is crucial as the original often appears distorted, as seen in the example below where I attempted to sketch Beauty and the Beast, particularly focusing on the female face.
The first run I got:
Then I did the upscale. I got this:
留言