Testing out Alibaba's new Z-image neural network.

The developers claim it as a compact, fast and open model for generating photorealistic images.

Plus, it's stated real-world knowledge (places, landmarks, people) and reasoning skills before drawing.

Simply put, it's like this. free Nano Banana, which you can run locally on your PC.

‍ Test results:

1. The model has figured out what the Eiffel Tower and Brooklyn Bridge look like, but for some reason the girl drinks coffee out of a cup in a cafe, but on a walk from a cup.
2. Photorealism, decent - the skin and fine details are there, but still needs an upscale.
3. Five fingers, all good.
4. The time, as not written, does not understand - neither 9.50, nor 15.30, and even more so without ten 10
5. Knows what Einstein and Bob Marley look like, but try to find Marilyn Monroe and Charlie Chaplin.

Testing for free here.