AI generators evolve before our eyes at a scary pace, but they still have flaws.
Spotting strange details in AI images is actually quite funny.
On closer inspection, the people’s weird hands gave them away as AI-generated images.
The fact that this was Midjourneys attempt at hands made the situation more interesting.
True enough, even DALL-E is prone to unrealistic fingers and nails.
Other engines are slow to follow Midjourneys example, sofixing AI art with Photoshopremains an invaluable skill.
The main hurdle for programmers is how complicated it is to train artificial intelligence to draw convincing hands.
Why Do AI Image Generators Struggle With Hands?
AI engines use generative adversarial networks (GANs) or Stable Diffusion to produce images.
Both technologies require extensive source materials, training, and processing power to create even the most basic artworks.
Whether straight or curled, showing five fingers or three.
Thats why Midjourney hands can be so ugly: AI confusion.
And their obstacles go beyond insufficient examples of hands.
But such achievements dont come cheap.
Training an AI to do better with hands requires feeding it better images, especially in 3D.
Even then, the software can make mistakes in otherwise stunning works of art.
Besides being a huge and complex job, its expensive.
So, dont expectfree AI text-to-image generatorsto step up to Midjourneys caliber just yet.
