Sign In

Yu_Person Perspective Close-up Enhancement LoRA for SD1.5

7

131

1

3

Updated: Aug 28, 2024

buildings

Verified:

SafeTensor

Type

LoRA

Stats

113

1

Reviews

Published

Aug 28, 2024

Base Model

SD 1.5

Trigger Words

person_perspective

Hash

AutoV2
01927DB90D
default creator card background decoration

kunkun008

Yu_Person Perspective Close-up Enhancement LoRA for SD1.5

Version 1.0 Introduction This model primarily enhances the effect of generating close-up architectural scenes from a person's perspective. It achieves the best results when used in combination with other 1.5 series models I've released.

Usage Suggestions

  1. When using, you can appropriately increase the CFG value; 5-9 is a reasonable range.

  2. Choose Euler a as the sampling method; 30 steps can achieve faster speed. To improve quality, you can choose DPM++3M Karras and set about 40 sampling steps.

  3. The generated image resolution should not be lower than 1024x1024, otherwise quality may decrease. This version of the model can achieve better results with high resolutions.

  4. It's recommended to use it with the 1.5 version base model I've released for the best effect. Link to the 1.5 version base model: https://civitai.com/models/622836/yu-architectural-rendering-model-sd15-v10

  5. It needs to be used with the accompanying prompts for the 1.5 base model, and "person_perspective" must appear for it to take effect.

  6. The recommended weight is 0.7 or above, can be set to 1, and can be used in combination with other models. When used in combination, please appropriately reduce the weight.

Version 0.8 Introduction: This model primarily enhances the effect of generating close-up architectural scenes from a person's perspective. It achieves the best results when used in combination with other 1.5 series models I've released.

Usage Suggestions

  1. When using, you can appropriately increase the CFG value; 5-9 is a reasonable range.

  2. Choose Euler a as the sampling method; 30 steps can achieve faster speed. To improve quality, you can choose DPM++3M Karras and set about 40 sampling steps.

  3. The generated image resolution should not be lower than 1024x1024, otherwise quality may decrease. This version of the model can achieve better results with resolutions close to 1024x1024.

  4. It's recommended to use it with the 1.5 version base model I've released for the best effect. Link to the 1.5 version base model: https://civitai.com/models/622836/yu-architectural-rendering-model-sd15-v10

  5. It needs to be used with the accompanying prompts for the 1.5 base model, and "person_perspective" must appear for it to take effect.

  6. The recommended weight is 0.7 or above, can be set to 1, and can be used in combination with other models. When used in combination, please appropriately reduce the weight.

1.0版本介绍

该模型主要加强人视点下生成近景建筑的效果

配套我发布的其他1.5系列模型使用才能达到最佳效果

使用建议

1.使用时可以适当拉高CFG值,5-9都是合理范围。

2.采样方法选择euler a,30步即可达到较快的速度;如需提高质量,可选择dpm++3m karras,并设置40左右的采样步数。

3.生成图片分辨率尽量不要低于1024x1024,否则可能会出现质量下降的情况。

该版本模型使用高分辨率可获得更好的效果

4.建议搭配我发布的1.5版本大模型使用效果最佳。

1.5版本大模型链接:https://civitai.com/models/622836/yu-architectural-rendering-model-sd15-v10

5.需结合1.5大模型的配套提示词,并且必须出现person_perspective才可生效。

6.权重建议为0.7以上,可以设为1,可与其他模型配合使用,配合使用时请适当降低权重。

0.8版本介绍:

该模型主要加强人视点下生成近景建筑的效果

配套我发布的其他1.5系列模型使用才能达到最佳效果

使用建议

1.使用时可以适当拉高CFG值,5-9都是合理范围。

2.采样方法选择euler a,30步即可达到较快的速度;如需提高质量,可选择dpm++3m karras,并设置40左右的采样步数。

3.生成图片分辨率尽量不要低于1024x1024,否则可能会出现质量下降的情况。

该版本模型使用接近1024x1024分辨率可获得更好的效果

4.建议搭配我发布的1.5版本大模型使用效果最佳。

1.5版本大模型链接:https://civitai.com/models/622836/yu-architectural-rendering-model-sd15-v10

5.需结合1.5大模型的配套提示词,并且必须出现person_perspective才可生效。

6.权重建议为0.7以上,可以设为1,可与其他模型配合使用,配合使用时请适当降低权重。