DogePoint 3D 游戏角色模型

使用介绍

此模型的工作原理与基本版本非常相似，只是默认情况下会剔除 2D 内容，并且一些烘焙角色的表现会更好。它不是 100% 完美，但已经是相当大的飞跃，仍然需要对 1.2 版进行一些调整，这就是您现在可能会发现的问题：修复伪影、某些采样器的升级效果不佳以及 score_9 的风格灵活性较低。

推荐的采样器/调度器：DPM++ 2M、Euler A / SGM Uniform、Normal

推荐CFG：5

推荐的upscalers: Latent / Tiled Upscalers

应该更加稳定，与 LoRA 更加兼容，同时保持风格并保留基础版本/PONY V6 的所有功能。

使用提示：

CLIP SKIP: 1

作品参照

<lora:MEME_Doge-PONY:0.8> kabosu, 1girl, upper body, huggigng shiba-inu plushie, yorha no.2 type b, blindfold, hairband, black dress, juliet gloves, abstract background, score_9, score_8_up, score_7_up, score_6_up,
Negative prompt: dogenegXL-64v-neg
Steps: 26, CFG scale: 5, Sampler: Euler a, Seed: 3445314925, VAE: fixFP16ErrorsSDXLLowerMemoryUse_v10.safetensors, Size: 512x768, Model: DogePONY-3D_V1.fp16, Version: f1.2.0dev-v1.10.1RC-latest-705-gdc1ba3958, TI hashes: [object Object], Model hash: 042ee63586, Hires steps: 18, Hires upscale: 1.75, Schedule type: SGM Uniform, Hires upscaler: Latent, dynthres_enabled: True, HyperTile_enabled: True, dynthres_cfg_mode: Constant, Denoising strength: 0.6, dynthres_sched_val: 1, HyperTile_max_depth: 0, HyperTile_swap_size: 2, HyperTile_tile_size: 256, dynthres_mimic_mode: Constant, dynthres_mimic_scale: 5, HyperTile_scale_depth: False, dynthres_cfg_scale_min: 0, dynthres_interpolate_phi: 1, dynthres_mimic_scale_min: 0, dynthres_scaling_startpoint: MEAN, dynthres_variability_measure: AD, dynthres_threshold_percentile: 1, dynthres_separate_feature_channels: enable, Clip skip: 2

tifa lockhart, solo, indoors, looking at viewer, leaning foward, bar counter, beverage bottles, cleavage, mismatching gloves, dark, score_9, score_8_up, score_7_up, score_6_up,
Negative prompt: dogenegXL-64v-neg
Steps: 26, CFG scale: 5, Sampler: Euler a, Seed: 844018690, VAE: fixFP16ErrorsSDXLLowerMemoryUse_v10.safetensors, Size: 512x768, Model: DogePONY-3D_V1.fp16, Version: f1.2.0dev-v1.10.1RC-latest-705-gdc1ba3958, TI hashes: [object Object], Model hash: 042ee63586, Hires steps: 18, Hires upscale: 1.75, Schedule type: SGM Uniform, Hires upscaler: Latent, ADetailer model: mediapipe_face_full, ADetailer steps: 20, dynthres_enabled: True, ADetailer sampler: Euler a, ADetailer version: 24.9.0, HyperTile_enabled: True, dynthres_cfg_mode: Constant, Denoising strength: 0.6, dynthres_sched_val: 1, ADetailer mask blur: 4, ADetailer scheduler: Karras, HyperTile_max_depth: 0, HyperTile_swap_size: 2, HyperTile_tile_size: 256, dynthres_mimic_mode: Constant, ADetailer confidence: 0.3, dynthres_mimic_scale: 5, HyperTile_scale_depth: False, ADetailer dilate erode: 4, dynthres_cfg_scale_min: 0, dynthres_interpolate_phi: 1, dynthres_mimic_scale_min: 0, ADetailer inpaint padding: 24, dynthres_scaling_startpoint: MEAN, ADetailer denoising strength: 0.4, ADetailer use separate steps: True, dynthres_variability_measure: AD, ADetailer inpaint only masked: True, dynthres_threshold_percentile: 1, ADetailer use separate sampler: True, dynthres_separate_feature_channels: enable, Clip skip: 2

matoi ryuuko:0.8, solo, short hair, streaked hair, bride veil, lace gloves, wedding dress, (holding katana, striking pose), red petals, blurred background, high angle, score_9, score_8_up, score_7_up, score_6_up,
Negative prompt: dogenegXL-64v-neg
Steps: 26, CFG scale: 5, Sampler: Euler a, Seed: 2121576639, VAE: fixFP16ErrorsSDXLLowerMemoryUse_v10.safetensors, Size: 768x1024, Model: DogePONY-3D_V1.fp16, Version: f1.2.0dev-v1.10.1RC-latest-705-gdc1ba3958, TI hashes: [object Object], Model hash: 042ee63586, Hires steps: 15, Hires upscale: 1.25, Schedule type: Karras, Hires upscaler: R-ESRGAN 4x+, ADetailer model: mediapipe_face_full, ADetailer steps: 20, ADetailer sampler: Euler a, ADetailer version: 24.9.0, HyperTile_enabled: True, Denoising strength: 0.5, ADetailer mask blur: 4, ADetailer scheduler: Karras, Hires schedule type: Karras, HyperTile_max_depth: 0, HyperTile_swap_size: 2, HyperTile_tile_size: 256, ADetailer confidence: 0.3, HyperTile_scale_depth: False, ADetailer dilate erode: 4, ADetailer inpaint padding: 24, ADetailer denoising strength: 0.4, ADetailer use separate steps: True, ADetailer inpaint only masked: True, ADetailer use separate sampler: True, Clip skip: 2