What's new:

I've added more training data focusing on back views and side views.

Due to some adjustments in the training parameters, the generalization is a bit limited this time.

For better results, try the following prompts :

For front view:

• foreshortening,view from below,footstomp

• splitt0es, foot focus, feet

For side view:

• low angle from side,worm's-eye view,foreshortening,footstomp

• s1t,leftvi3w, crossed legs,feet,5toes,sitting（1：1）

• s1t,leftvi3w, crossed legs,feet, barefoot,sitting（4：3）

• s1t,leftvi3w, crossed legs,feet,5toes,sitting（3：4）

For back view:

• lying down,lying on bed

• 1iedown,backvi3w, back view, soles（1：1）

• 1iedown, backvi3w, back view, 5toes（4：3）

• 1iedown,lying down, backvi3w, back view, soles（3：4）

You might find that the model responds differently depending on the canvas size.

I highly recommend experimenting with different prompt combinations—especially when switching between various aspect ratios—to find the best results for your specific composition.