Update README.md
Browse files
README.md
CHANGED
|
@@ -3,6 +3,8 @@
|
|
| 3 |
---三个融合模型，1.0偏向2.5d，3.0是适合搭配高清修复绘制1k大图的2d模型，小图效果一般，2.0则是二者之间，比较全面，建议配合”deepnegative”embedding使用，使用时tag可以尽量简洁，推荐参数：Steps: 28, Sampler: DPM++ SDE Karras, CFG scale: 8, Denoising strength: 0.56, Clip skip: 2, ENSD: 31337, Hires upscale: 2, Hires steps: 20, Hires upscaler: Latent (nearest-exact), Eta: 0.67。
|
| 4 |
注意，以上参数以外的参数我并没有尝试过，效果无法保证。只推荐vae-ft-mse-840000-ema-pruned作为vae
|
| 5 |
|
|
|
|
|
|
|
| 6 |
|
| 7 |
|
| 8 |
以下预览图皆使用了deepnegative
|
|
|
|
| 3 |
---三个融合模型，1.0偏向2.5d，3.0是适合搭配高清修复绘制1k大图的2d模型，小图效果一般，2.0则是二者之间，比较全面，建议配合”deepnegative”embedding使用，使用时tag可以尽量简洁，推荐参数：Steps: 28, Sampler: DPM++ SDE Karras, CFG scale: 8, Denoising strength: 0.56, Clip skip: 2, ENSD: 31337, Hires upscale: 2, Hires steps: 20, Hires upscaler: Latent (nearest-exact), Eta: 0.67。
|
| 4 |
注意，以上参数以外的参数我并没有尝试过，效果无法保证。只推荐vae-ft-mse-840000-ema-pruned作为vae
|
| 5 |
|
| 6 |
+
Three fusion models, 1.0 biased towards 2.5d, 3.0 is a 2d model suitable for high-definition repair and drawing 1k large images. The small image effect is average, and 2.0 is between the two. It is recommended to use with "deep negative" embedding. When using, the tag can be as simple as possible. Recommended parameters: Steps: 28, Sample: DPM++SDE Karras, CFG scale: 8, Denoising strength: 0.56, Clip skip: 2, ENSD: 31337, Hires upscale: 2, Hires steps: 20, Hires upscaler: Latent (nearest-exact), Eta: 0.67。 Note that I have not tried parameters other than the above parameters, and the effect cannot be guaranteed. Only vae-ft-mse-840000-ema-pruned is recommended as vae
|
| 7 |
+
|
| 8 |
|
| 9 |
|
| 10 |
以下预览图皆使用了deepnegative
|