Merge of Nova Furry V4, Illustrious V1.0(paid for), and WaiV10 Nova Furry V4 ~ <a target="_blank" rel="ugc" href="https://civarchive.com/models/503815?modelVersionId=1402403">https://civarchive.com/models/503815?modelVersionId=1402403</a> Illustrious V1.0 ~ <a target="_blank" rel="ugc" href="https://civarchive.com/models/1232765/illustrious-xl-10">https://civarchive.com/models/1232765/illustrious-xl-10</a> WaiV10 ~ <a target="_blank" rel="ugc" href="https://civarchive.com/models/827184?modelVersionId=1367422">https://civarchive.com/models/827184?modelVersionId=1367422</a> I merged the models using dare mergers Comfy UI, i almost don't want to share it but it's really good that i had to share it. use Euler A/DPMPP_SDE or other variant for best result.use Karras for best result CFG: 2.5 ~ 4.5 (Lower better quality but less adhered to prompt. Higher more adhered to prompt degraded quality) Steps: 10 ~ 40 (higher steps might be required to make the models adhered to prompts but it doesn't necessarily increase quality although i can see some difference in details.) VAE and CLIP integrated. Note to use original CLIP from checkpoint, using other checkpoint CLIP might be unstable and yield bad results. You can also use SDXL VAE.1024x1024 up to 1536x1536(very unstable/bad results). 1248x1824 max resolution(stable/good results). The models will follow your prompt quite literally might results in fusion or generalization for multiple subject. Make sure to add solo, duo, trio, ETC for multiple subject.The first prompt will be weighted more so make sure to put your prompt in order of priority. The only solution i've found is adding weight to prompt like. (catgirl:1.2) something like that. Also don't put more than weight 1.8 per prompt might result in degraded quality.For LORA use 0.8 weight so the models still have room for creativity and details. The quality depends on your LORA too. IP-Adapter can be used for style transfer. Use 0.8 weight for IP-Adapter. Use Red mask for attn-masking.For img2img encode your image using VAE then put the latent to sampler and use 0.65 ~ 0.85 denoise on sampler. 2.5 ~ 3 CFG and 20~40 steps. Negative Prompt:<pre><code>neg: lowres, worst quality, low quality, bad anatomy, bad proportions, fused fingers, missing fingers, watermark, username, censored, unclear eyes, blurry eyes, ai-generated, lowres, jpeg artifacts, old, oldest, Base:bad quality,worst detail,sketch,censor,</code></pre> Honorable mention @Crody Nova Furry Creator @ONOMAAI Illustrious XL V1.0 @WAI0731 Wai V10 Creator Let me know if there anyway i can improve this cheers...

Files with SHA256: d06b4cca8374914a3e358dffe229aca6e1c6dc7dedad1c92eac1120948ffb94c