Niji v6 testing. I trained this alongside with niji jelly so you can use the same trigger word too lol. The only difference? This is more refined than niji jelly, I planned to make it a separate dataset but I felt lazy so I just retrained it with the same dataset. Overall it doesn't need any trigger words honestly, as long as you mention pseudo realistic.Strength usage: 1 or 1.2 as a base. It's a bit more blurry/airbrushed. If you use 0.8 it'll be more similar to niji jelly style just more cartoonish.I recommend using an extra detailer support, or any lora that'll help lessen the blurry lines since it's way too tough looking depending on the generation, but overall to me it doesn't need to. I'll only be showcasing the gens made with this lora as a base.Unlike niji jelly, shorter prompts are alright. But longer ones are definitely better.This has better backgrounds, better lighting. The only problem? Although it has fairly good hands sometimes it'll look blurry. The jewelries are slightly deformed but it's not noticeable honestly. Only because this is more of an oil painting vibe.Sampling steps: 8-12. 8 is fine and it's cheaper, but you can do higher for more detail. I haven't tried higher yet since I'm currently saving credits while testing another lora.If you're curious about niji jelly, this one is the softer version of niji classic. Both are trained from niji v6. I had to relearn how the hell to use niji again to get the dataset. At least my old gens helped lol<a target="_blank" rel="noopener noreferrer nofollow" href="https://tensor.art/models/1027003717422982293?source_id=kDmwol7ilEexpfQjaX32-hch">Link here: nijijelly</a>

Files with SHA256: 7d9dd4e5c44ed112e340b8cc1c8e42d2b193b7a449e43998189d5913d9555c64