Used to hunt for models through scattered channels, versions never matched and sample images were false advertising half the time. Now I mainly use a model community, models and LoRAs are centralized, come with sample images and parameter references, one glance and you roughly know the output style, way less trial and error.
For someone like me who cares more about “can I reproduce this,” the trigger words and recommended weights authors attach are super useful, saves me from blindly guessing.
Downside is heavy homogenization among popular models, half the images in the community look the same flavor. When you pick a LoRA do you go mainly by sample images or download count?