Sign In

中国电动汽车Cars | China EV_V1.1

182
1.4k
12
Verified:
SafeTensor
Type
LoRA
Stats
1,065
Reviews
Published
Jun 25, 2023
Base Model
SD 1.5
Trigger Words
China EV
CNEV
Hash
AutoV2
D97530A162
default creator card background decoration
cliffcat's Avatar
cliffcat

基于中国的电动汽车外形训练的模型,可以输出【封闭式中网,隐藏式门把手,无排气管】的电动汽车外观,给汽车设计提供更多可落地的设计灵感。

汽车爱好者欢迎一起交流,返图、评论、打分及收藏。

The model based on the shape training of electric vehicles in China can output the appearance of electric vehicles with closed center net, hidden door handle and no exhaust pipe, providing more landing design inspiration for car design.

Car enthusiasts are welcome to exchange, rate, collect, actively return pictures and comments.

有任何问题,可以抖音私信我 “隐公子E车

···················································································································································

如果您喜欢这个模型,请考虑给我一杯咖啡

If you like the model, please consider tipping me a coffee.

···················································································································································

V1.1

新版使用38款车共300多张图训练而成,V1.1累计24炉共测试生成3万多张图,炼丹不易请多多返图!!!

五分制多底模测试得分,V1.1平均分比1.0高0.065分。

  1. 增加5台新车的样式,AI辅助你碰出更多火花!

  2. 增加竖图2:3(建议1car加权重),正方1:1 的支持;

  3. 增加噪声偏移,白更亮,黑更暗,画面色彩更纯净;

  4. 修复black car不出全黑车型的问题;

  5. 修复汽车前脸不对称,包含车灯、日行灯、保险杠等。

彩蛋:车灯特写,竖图激活多抽卡。

The new version is trained using over 300 images from 38 vehicles, and V1.1 has accumulated 24 rounds of testing to generate over 30000 images. Training is not easy, please return more images!!!

The score of the five point multi base model test is 0.065 points higher than the average score of 1.0 in V1.1.

  1. Add 5 new car styles with AI assistance to create more sparks!

  2. Add a vertical image of 2:3 (recommended weighting of 1 car), with support for a 1:1 square;

  3. Increase noise offset, make white brighter, black darker, and the color of the image purer;

  4. Fix the issue of black cars not producing all black models;

  5. Fix the asymmetry of the car's front face, including headlights, daytime running lights, bumpers, etc.

Egg: Close up of car headlights, vertical image activation with multiple card draws.

---------------------------------

出图说明:

触发词Trigger Words:China EV,CNEV(竖图)

推荐值

  • CLIP 终止层数:1

  • 权重Weight:0.6-0.8,最大不要超过0.9

  • 步数Steps:20~30

  • 尺寸Px: 768*512 (3:2)、512*768 (2:3)、512*512 (1:1)

  • 采样Sampler:DPM++2MKarras

其他说明:

  • 出图默认以轿车为主(Default is mainly sedan),

  • 运动多功能车型 填写“SUV”并增加↑权重Weight;

  • 小型车 填写“small car”并增加↑权重Weight;

  • 跑车 填写“sports car”或 跑车品牌,例如Such as“Ferrari, Lamborghini, Porsche”。

与以下底模、采样配合出图较好:

The graph is well matched with the following models and samples:

【底模models】

Chilloutmix-fp32-fix,Deliberate,Henmixreal_v30,Perfect world_v4,Majicmixreal_v4,RealisticVisionV20,GhostmixV20,Mixreal_v2,Unstableinkdream,Comix_v3,Dvauto_v1,Anything_v5,Knollingcase,3guofeng3_v33,Nigi3d_v20,SavMIX11,

【采样Sampler】

DPM++ 2M SDE Karras,DPM adaptive,LMS Karras,DPM2 a Karras,DPM++2S a Karras,DPM++ 2M Karras,DPM++ SDE Karras,UniPC

---------------------------------

写实类推荐: Chilloutmix

Q版卡通类推荐:Nigi3d_v20

---------------------------------

XYZ 高清图 移步 https://www.123pan.com/s/yP48Vv-meHHd.html 提取码:CNEV