>>72908895A few more onsen pics of pekomama and Biboo.
https://files.catbox.moe/99sq21.pnghttps://files.catbox.moe/l7nbtz.pnghttps://files.catbox.moe/35gr0u.pngSome bonus pics from when I tested the new models with some old workflows.
Ina:
https://files.catbox.moe/sso088.pngRaden:
https://files.catbox.moe/izo60t.pngLooking forward to redoing many of my old images, but I'll call it a night for now.
>>72909264I guess it depends heavily on what you like, but it seems like some pony loras also work with animagine 3.1. At least, they do for
>>72897908this anon.
>>72909394I knew I wasn't imagining things. One of my old prompts had "from above" for the background, which previously just slightly lifted the camera sometimes. Now, it's giving me shots of the entire city.
And it's honestly kind of crazy how good they are at inpainting by themselves (eg:
https://files.catbox.moe/k90g5a.png). While actual inpainting models are still better, it's kind of crazy. I mean, it's still just SDXL, but it's such a leap in quality. Almost feels like we got new base models (which honestly makes me very excited for SD3.0, and what that will be able to do after a year of finetunes)
>>72912280Yeah, I saw those Okayu/Mio gens that were done with TCD a few threads back, and that's just mindblowing. Making any model a Turbo model(I guess they are called Lightning now?) just by using a Lora. Sure, it's a bit less refined, but that's super handy for quick testing/getting a selection of images.
Also, very pretty gen, anon.
>>72912621Honestly, I feel I could do better, but this may be what that other anon meant with the new models needing different prompting:
https://files.catbox.moe/f2e9vp.pnghttps://files.catbox.moe/p6fbze.pnghttps://files.catbox.moe/mdvj29.pnghttps://files.catbox.moe/z6sttb.png>>72915089Honestly, I feel like I returned at just the right time. Also, those are some amazing images. Maybe I need to look into forge if it can do something like that.