Presumably from last year until now, everyone knows more and more about AI, and I also sighed at first, finally! Finally! Everyone began to pay attention to how to apply AI to the design. I was afraid that the teacher would find out that my design was generated by AI. Now I can sort out the relevant contents and share them with my colleagues in Niu Niu. Here I will use a car robot case to share with you what can be achieved with AI.

Production time: 4 days

Omitting numerous prefixes, we began to train LORA after the concept was finalized. Its essence is to let the existing FLUX model (one of the currently relatively easy-to-use open source models) know what a car robot is, what its structure is, whether it has a pan-tilt, look for many reference diagrams here, and then mark it together with our intended modeling, etc. for training.

Yunbao, which has both the softness of a cloud and the silhouette shape of a silver ingot, has been modified by repeated AI and its shape has been determined.

Let's add a little detail, here we can use ps, p good map, then directly throw to our training model, fusion, of course, remember Mr. into the back view and find the desired texture modeling.

I don't know whether to add to the picture or enrich the details. When I did this, I simply gave a few interesting expressions. I also wanted Party A to look more vivid and have a sense of picture.

The second shape, which I think is good, is a very concrete representation of the concept of clouds.

Try to analyze its material, as well as modeling

Here is a supplement to the three views, which enables our designers to give the team and Party A specific forms in the early stage of design. Of course, it is very fast, and it is also convenient for us to quickly connect the concept scheme.

Here is also divergent thinking, not limited to clouds, but from the perspective of bionics, the output of two I think a little interesting modeling divergence.