Thanks for the LoRAs!

For some reason, v2 seems to be creating noisier results compared to v1. It could just be my setup though!

v1: sample 1, sample 2

v2: sample 1, sample 2

Reducing the weight or switching the sampler to the one in the examples helps, but it's strange that only v2 is doing this. Everything was kept the same aside from the LoRA and trigger word. Here's a simpler prompt using the sampler in the examples (v1, v2). The sampler does smooth the noise out into "details" but I think v1 is still cleaner.