The description of this LoRA is a single-sentence tragedy on the level of "For sale: baby shoes, never worn"

After further testing, my initial hypothesis regarding native resolution training didn't fully hold up. Some recent renders showed, that this LoRA is nearly impossible to mix with other character LoRAs because the silhouette is ultra-dominant. This strongly suggests a very high Rank (Dim) configuration, far exceeding the average Krea2/Flux models on Civitai. However, as the author mentioned, it was trained specifically as an avatar model—so this extreme dominance and high-dim rigidity make sense as a feature to enforce strict character consistency. For his purpose - a dream result.