最近找了阿卡丽的Lora，因为动漫checkpoint模型的色彩元素比较丰富，目标是生成真人版阿卡丽奔跑攻击的动作，所以就

1，通过prompt抽奖，先生成分辨率尚可的960*540动漫版本的原始图素，

2，通过过controlnet将depth作为输入通道不固定色域只是固定大致动作继而生成最终的图片

prompt:realistic,in the open air,masterpiece,ultra detailed,looking looking at another,attack,blonde and black hair, graceful,cute,(full body,feet,akali,highres, <lora:kda_all_out_akali_v1:0.7>, 1girl, k/da \(league of legends\), beautiful eyes,beautiful face, ), sexy:0.1,holographic interface,<lora:add_detail:0.7>,<lora:fashigirl-v5.5-lora-naivae-64dim:0.4> , scampering with sickle:1.3,nudist beach uniform,brown eyes,Negative prompt: verybadimagenegative_v1.3, NegfeetV2, ng_deepnegative_v1_75t, bad_prompt_version2, bad-image-v2-39000, FastNegativeV2,( EasyNegative:0.6), bad-hands-5,split,pieces,low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry, lowres graffiti, (low quality lowres simple background:1.1),lying down,Steps: 60, Sampler: DDIM, CFG scale: 7, Seed: 1501575983, Face restoration: GFPGAN, Size: 3840x2160, Model hash: 18ed2b6c48, Model: xxmix9realistic_v40, Denoising strength: 0.2, Clip skip: 2, Tiled Diffusion upscaler: R-ESRGAN 4x+, Tiled Diffusion scale factor: 2, Tiled Diffusion: {"Method": "MultiDiffusion", "Tile tile width": 160, "Tile tile height": 96, "Tile Overlap": 88, "Tile batch size": 1, "Upscaler": "R-ESRGAN 4x+", "Upscale factor": 2, "Keep input size": true}, ControlNet 0: "preprocessor: tile_resample, model: control_v11f1e_sd15_tile [a371b31b], weight: 1, starting/ending: (0.5, 1), resize mode: Crop and Resize, pixel perfect: True, control mode: ControlNet is more important, preprocessor params: (512, 1, 64)", ControlNet 1: "preprocessor: depth_leres, model: control_v11f1p_sd15_depth [cfd03158], weight: 1, starting/ending: (0, 1), resize mode: Crop and Resize, pixel perfect: True, control mode: Balanced, preprocessor params: (512, 0, 0)", Lora hashes: "kda_all_out_akali_v1: 5515093a58db, add_detail: 7c6bad76eb54, fashigirl-v5.5-lora-naivae-64dim: 2dda5ffaa304", TI hashes: "verybadimagenegative_v1.3: d70463f87042, NegfeetV2: df90b1ff666d, ng_deepnegative_v1_75t: 54e7e4826d53, bad_prompt_version2: 6f35e7dd816a, bad-image-v2-39000: 5b9281d7c6ad, FastNegativeV2: a7465e7cc2a2, EasyNegative: c74b4e810b03, bad-hands-5: aa7651be154c", Version: v1.5.1

controlnet depth图生图成品

rev animation中间态大模型

初始模型flat2DAnimerge