SDXL
SDXL
Cat Tower is anime style NoobAI XL checkpoint.VAE is baked in.v-prediction versions are experimental models.You need to use the webui that support v-prediction.ComfyUI (recommended)Forge NEO / Classic (recommended)reForgeForge (maybe)AUTOMATIC1111 (from v1.4, and you need to place the config file in the same folder as the checkpoint)LicenseThis model adheres to the NoobAI-XL License (Modified Fair AI Public License 1.0-SD) and is available for non-commercial use only. This license prohibits any form of commercialization, including but not limited to monetization or commercial use of the model, derivative models, or model-generated products.Recommended SettingSteps: 25-30CFG scale: 5-7Sampler (vPred): I recommend using "Euler". Using "Euler a" sometimes generates colored skin and objects such as red, blue and so on.Sampler (Epsilon-pred): DPM++ 2M Karras, Euler aScheduler (vPred) : Normal, Simple ("Automatic" produce noisy images.)Use ADetailer as neededPositive Promptmasterpiece, best qualityNegative Promptworst quality, bad quality, low quality, lowres, scan artifacts, jpeg artifacts, sketch, light particlesMerge recipes (ChenkinNoob v1.2)Full Finetuning ChenkinNoob-XL-V0.5 with anime style dataset(2.1k), optimizer = AdamW8bitKahan, scheduler = cosine, learning rate = 2.5e-6, 30 epoch, unet only, batch size = 2, gradient accumulation steps = 8. (Checkpoint A)Merging Cat Carrier v8.0 to Checkpoint A by "Weighted Sum" with block (0,0,0,0,0,0,0.5,0.5,0.5,0.5,0.5,0.5,0.5,0.5,0.5,0,0,0,0,0) (Checkpoint B)Merging the difference between copycat-RouWei vpred-v0.42 and Rouwei 0.80 vpred to Checkpoint B by "Train Difference" with block (0,0,0,0,0,0,0,0,0,0,0,0.3,0.3,0,0,0,0,0,0,0) (Cat Tower ChenkinNoob v1.2)Merge recipes (ChenkinNoob v1.1)Merging the difference between ChenkinNoob-XL-V0.2 and stable-diffusion-xl-base-1.0, CyberRealistic XL v7.0 and stable-diffusion-xl-base-1.0 to ChenkinNoob-XL-V0.2 by "Perpendicular Component" and "Add Difference" (Checkpoint A)Merging Checkpoint A to ChenkinNoob-XL-V0.2 by "SLERP", alpha = 0.5 (Checkpoint B)Merging Checkpoint B to ChenkinNoob-XL-V0.2 by "Rotate", alignment = 1.0, alpha = 0.0 (Checkpoint C)Replacing the CLIP of Checkpoint C by ChenkinNoob-XL-V0.2 (Checkpoint D)Finetuning Checkpoint D with anime style dataset(0.5k) by OFT method, dim = 4, alpha = 1e-3, learning rate = 2.5e-6, 15000 steps. And merging the OFT model to ChenkinNoob-XL-V0.2, ratio = 1.0 (Checkpoint E)Merging Checkpoint E to Checkpoint D by "SLERP" with block (0,0,0,0,0,0,0,0,0,0,0,0.7,0.7,0.7,0.7,0,0,0,0,0) (Checkpoint F)Merging Cat Carrier v8.0 to Checkpoint F by "Weighted Sum" with block (0,0,0,0,0,0,0.5,0.5,0.5,0.5,0.5,0.5,0.5,0.5,0.5,0,0,0,0,0) (Checkpoint G)Merging the difference between Cat Carrier v8.0 and Raehoshi illust XL v7.1 to Checkpoint G by "Add Difference" with block (0,0,0,0,0,0,0,0,0,0,0,1,1,0,0,0,0,0,0,0) (Checkpoint H)Merging the difference between LunarPeachMix v2.0 and Illustrious XL v1.1 to Checkpoint H by "Train Difference" with block (0,0,0,0,0,0,0,0,0,0,0,0.3,0.3,0,0,0,0,0,0,0) (Checkpoint I)Merging the difference between copycat-RouWei vpred-v0.42 and Rouwei 0.80 vpred to Checkpoint I by "Train Difference" with block (0,0,0,0,0,0,0,0,0,0,0,0.3,0.3,0,0,0,0,0,0,0) (Checkpoint J)Merging style LoRA to Checkpoint J (Cat Tower ChenkinNoob v1.1)Merge recipes (ChenkinNoob v1.0)Merging the difference between ChenkinNoob-XL-V0.1 and stable-diffusion-xl-base-1.0, CyberRealistic XL v7.0 and stable-diffusion-xl-base-1.0 to ChenkinNoob-XL-V0.1 by "Perpendicular Component" and "Add Difference" (Checkpoint A)Merging Checkpoint A to ChenkinNoob-XL-V0.1 by "SLERP", alpha = 0.5 (Checkpoint B)Merging Checkpoint B to ChenkinNoob-XL-V0.1 by "Rotate", alignment = 1.0, alpha = 0.0 (Checkpoint C)Replacing the CLIP of Checkpoint C by ChenkinNoob-XL-V0.1 (Checkpoint D)Finetuning Checkpoint D with anime style dataset(0.5k) by OFT method, dim = 4, alpha = 1e-3, learning rate = 2.5e-6, 15000 steps. And merging the OFT model to ChenkinNoob-XL-V0.1, ratio = 1.0 (Checkpoint E)Merging Checkpoint E to Checkpoint D by "SLERP" with block (0,0,0,0,0,0,0,0,0,0,0,0.7,0.7,0.7,0.7,0,0,0,0,0) (Checkpoint F)Merging Cat Carrier v8.0 to Checkpoint F by "Weighted Sum" with block (0,0,0,0,0,0,0.5,0.5,0.5,0.5,0.5,0.5,0.5,0.5,0.5,0,0,0,0,0) (Checkpoint G)Merging the difference between Cat Carrier v8.0 and Raehoshi illust XL v7.1 to Checkpoint G by "Add Difference" with block (0,0,0,0,0,0,0,0,0,0,0,1,1,0,0,0,0,0,0,0) (Checkpoint H)Merging the difference between LunarPeachMix v2.0 and Illustrious XL v1.1 to Checkpoint H by "Train Difference" with block (0,0,0,0,0,0,0,0,0,0,0,0.3,0.3,0,0,0,0,0,0,0) (Checkpoint I)Merging the difference between copycat-RouWei vpred-v0.42 and Rouwei 0.80 vpred to Checkpoint I by "Train Difference" with block (0,0,0,0,0,0,0,0,0,0,0,0.3,0.3,0,0,0,0,0,0,0) (Checkpoint J)Merging style LoRA to Checkpoint J (Cat Tower ChenkinNoob v1.0)Merge recipes (Obsession vPred v2.0)Merging the difference between Obsession (Illustrious-XL) v-pred_v2.0 and stable-diffusion-xl-base-1.0, CyberRealistic XL v5.6 and stable-diffusion-xl-base-1.0 to Obsession (Illustrious-XL) v-pred_v2.0 by "Perpendicular Component" and "Add Difference" (Checkpoint A)Merging Checkpoint A to Obsession (Illustrious-XL) v-pred_v2.0 by "SLERP", alpha = 0.5 (Checkpoint B)Merging Checkpoint B to Obsession (Illustrious-XL) v-pred_v2.0 by "Rotate", alignment = 1.0, alpha = 0.0 (Checkpoint C)Replacing the CLIP of Checkpoint C by Obsession (Illustrious-XL) v-pred_v2.0 (Checkpoint D)Finetuning Checkpoint D with anime style dataset A(2.1k) by full finetune, optimizer = AdamW8bit, scheduler = warmup stable decay, learning rate = 6e-6, scheduler min lr ratio = 0.1, 32 epoch, unet only, gradient accumulation steps = 1, batch size = 4. (Checkpoint E)Finetuning Checkpoint D with anime style dataset B(0.5k) by full finetune, optimizer = AdamW8bit, scheduler = warmup stable decay, learning rate = 3e-6, scheduler min lr ratio = 0.1, 60 epoch, unet only, gradient accumulation steps = 1, batch size = 4. (Checkpoint F)Merging Checkpoint F to Checkpoint E by "SLERP", alpha = 0.5 (Checkpoint G)Merging the difference between LunarPeachMix v2.0 and Illustrious XL v1.1 to Checkpoint G by "Add Difference" with block (0,0,0,0,0,0,0,0,0,0,0,1,1,1,1,0,0,0,0,0) (Checkpoint H)Merging the difference between JANKU v5.0 and RouWei v0.8 epsilon to Checkpoint H by "Add Difference" with block (0,0,0,0,0,0,0,0,0,0,0,1,1,1,1,0,0,0,0,0) (Checkpoint I)Merging the difference between copycat-RouWei vpred-v0.42 and Rouwei 0.80 vpred to Checkpoint I by "Add Difference" with block (0,0,0,0,0,0,0,0,0,0,0,1,1,1,1,0,0,0,0,0) (Checkpoint J)Merging the Checkpoint I by "Clamp", bounds = Checkpoint G, LunarPeachMix v2.0, JANKU v5.0 and copycat-RouWei vpred-v0.42 (Checkpoint K)Add v_pred and ztsnr keys to Checkpoint K by noob_v_pencil-XL merge recipes (Cat Tower vPred v2.0)Merge recipes (Epsilon-pred v1.4)Merging the difference between NoobAI-XL Epsilon-pred v1.1 and stable-diffusion-xl-base-1.0, CyberRealistic XL v5.6 and stable-diffusion-xl-base-1.0 to NoobAI-XL Epsilon-pred v1.1 by "Perpendicular Component" and "Add Difference" (Checkpoint A)Merging Checkpoint A to NoobAI-XL Epsilon-pred v1.1 by "SLERP", alpha = 0.5 (Checkpoint B)Merging Checkpoint B to NoobAI-XL Epsilon-pred v1.1 by "Rotate", alignment = 1.0, alpha = 0.0 (Checkpoint C)Replacing the CLIP of Checkpoint C by NoobAI-XL Epsilon-pred v1.1 (Checkpoint D)Finetuning Checkpoint D with anime style dataset(0.5k) by OFT method, dim = 4, alpha = 1e-3, learning rate = 2.5e-6, 15000 steps. And merging the OFT model to Checkpoint D, ratio = 1.0 (Checkpoint E)Merging Cat Carrier v6.0 to Checkpoint E by "Weighted Sum" with block (0,0,0,0,0,0,0.5,0.5,0.5,0.5,0.5,0.5,0.5,0.5,0.5,0,0,0,0,0) (Checkpoint F)Merging the difference between LunarPeachMix v2.0 and Illustrious XL v1.1 to Checkpoint F by "Add Difference" with block (0,0,0,0,0,0,0,0,0,0,0,1,1,1,1,0,0,0,0,0) (Checkpoint G)Merging the difference between JANKU v4.0 and RouWei v0.7 epred to Checkpoint G by "Add Difference" with block (0,0,0,0,0,0,0,0,0,0,0,1,1,1,1,0,0,0,0,0) (Checkpoint H)Merging the difference between copycat-RouWei vpred-v0.42 and Rouwei 0.80 vpred to Checkpoint H by "Add Difference" with block (0,0,0,0,0,0,0,0,0,0,0,1,1,1,1,0,0,0,0,0) (Checkpoint I)Merging the Checkpoint I by "Clamp", bounds = Checkpoint F, LunarPeachMix v2.0, JANKU v4.0 and copycat-RouWei vpred-v0.42 (Cat Tower Epsilon-pred v1.4)Merge recipes (vPred v1.8)Merging the difference between NoobAI-V-Pred-1.0-Version and stable-diffusion-xl-base-1.0, CyberRealistic XL v5.6 and stable-diffusion-xl-base-1.0 to NoobAI-V-Pred-1.0-Version by "Perpendicular Component" and "Add Difference" (Checkpoint A)Merging Checkpoint A to NoobAI-V-Pred-1.0-Version by "SLERP", alpha = 0.5 (Checkpoint B)Merging Checkpoint B to NoobAI-V-Pred-1.0-Version by "Rotate", alignment = 1.0, alpha = 0.0 (Checkpoint C)Replacing the CLIP of Checkpoint C by NoobAI-V-Pred-1.0-Version (Checkpoint D)Finetuning Checkpoint D with anime style dataset(0.5k) by OFT method, dim = 4, alpha = 1e-3, learning rate = 2.5e-6, 15000 steps. And merging the OFT model to Checkpoint D, ratio = 1.0 (Checkpoint E)Merging Cat Carrier v6.0 to Checkpoint E by "Weighted Sum" with block (0,0,0,0,0,0,0.5,0.5,0.5,0.5,0.5,0.5,0.5,0.5,0.5,0,0,0,0,0) (Checkpoint F)Merging the difference between LunarPeachMix v2.0 and Illustrious XL v1.1 to Checkpoint F by "Add Difference" with block (0,0,0,0,0,0,0,0,0,0,0,1,1,1,1,0,0,0,0,0) (Checkpoint G)Merging the difference between JANKU v4.0 and RouWei v0.7 epred to Checkpoint G by "Add Difference" with block (0,0,0,0,0,0,0,0,0,0,0,1,1,1,1,0,0,0,0,0) (Checkpoint H)Merging the difference between copycat-RouWei vpred-v0.42 and Rouwei 0.80 vpred to Checkpoint H by "Add Difference" with block (0,0,0,0,0,0,0,0,0,0,0,1,1,1,1,0,0,0,0,0) (Checkpoint I)Merging the Checkpoint I by "Clamp", bounds = Checkpoint F, LunarPeachMix v2.0, JANKU v4.0 and copycat-RouWei vpred-v0.42 (Checkpoint J)Add v_pred and ztsnr keys to Checkpoint J by noob_v_pencil-XL merge recipes (Cat Tower vPred v1.8)Merge recipes (Epsilon-pred v1.3)Merging Cat Tower Epsilon-pred v1.2 and Cat Tower v-pred v1.7 by "DARE-TI