This requires minimal work to achieve equal or above his excellent samples. Make sure to use "meinapastel" checkpoint v4 or v6 (I like v6 better) to get started, then try others. Many .safetensor models look awful. Also use an inspector (google it) to see prompt and negative prompts for his samples. Lower the Lora weight far below default "1" to somewhere between .25 and .5. I also recommend using slight sharpening and stronger detail textual inversions/loras.

It seems more oriented for landscape dimensions. Portraits seemed slightly less vibrant and creative. YMMV.