With this workflow you will be able to generate a 3x3 grid with the same character face in 9 different poses and with small expression differences.
The workflow will output an upscaled image that can then be split in 9 different square image for LoRA training.
