SilvermoonMix03 focuses natural lighting, performing exceptionally well in low-light environments and high-contrast scenarios. It is specifically tuned for abstract elements and fluid textures, such as hair, water, ribbons, etc.
For instructions on how to use the V-pred model, please refer to https://civarchive.com/models/833294/noobai-xl-nai-xl
Recommended settings:
Sampling method: Euler a
Start with no negative prompt. Sample images were generated without negative prompt.
CFG: 4-5
Use ADetailer extension to enhance facial details: https://github.com/Bing-su/adetailer
推荐设置：
采样方法：Euler a
先尝试不使用负面提示词。示例图片都没有使用负面提示词。
CFG: 4.5-5.5
使用 ADetailer 扩展来增强面部细节：https://github.com/Bing-su/adetailer
Change notes:
v6.0: Significantly enhanced lighting, delivering even more natural lighting than v5.0.
v5.0: Lighting looks more nature now. Improved logic.
v4.0: Improved compatibility with LoRAs that are trained on other models.
v3.1: Girls now look older (longer face and limbs). Adjusted shadows in low light images.
v3.0: Now supports dark images
v2.0: Updated training strategy improves all areas, including but not limited to: hands, feet, background, styles
v1.1: Skin texture improvements
v1.0: Initial release
Description
FAQ
Comments (11)
can i have your wan workflow . your clip look so smooth i want to learn how you can do that pls
The key is to generate 24 FPS video and then use frame interpolation to increase it to 48 FPS.
https://civitai.com/models/1824577/img-to-video-simple-workflow-wan22-or-gguf-or-lora-or-upscale-or-fast
@silvermoong how many vram did you use and the size off image , i have 16gb vram but it keep said im out of vram even when using q4ks gguf model to run 512x512 + upscale
I have no idea why, but this model seems to REALLY like to add more than 1 person when generating (almost always somewhere in the background), even when using 1girl,solo. No other model i've tested with the same prompts have the same issue (at least not as strongly)...
@richerich Could you please share me your prompt with this issue?
@silvermoong basically any prompt if they are similar enough, like woman,female,girl in any other format...
Example: young woman,dark-skinned female,catgirl almost always produces 2 or more characters, even with the other prompts present. At least for me.
@richerich I just rolled a few images with woman,dark-skinned female,catgirl (I removed "young" to avoid TOS), I'm not seeing this issue: https://civitai.com/posts/24694327
Images generated by WebUI, you can right-click ⇾ open link in new tab to download the original images with generation metadata.
@silvermoong Yeah, i get that. However, as you can see in the images i uploaded, as soon as you start increasing the strength by any amount of said prompts, you will start seeing 2 or more characters. Other models also do this, but not nearly as early or strongly. That's what i'm hinting at. I have to increase the 1girl,solo to match the other prompts with this model, while other models dont need that until at least 1:3 strength: https://civitai.com/posts/24695761
@richerich Hi, I saw your images. The issue comes from ((mature woman, lady, madam):1.3).
My model follows prompts precisely. However, these prompts tend to generate more characters rather than changing the main character.
Check my example of adjusting the main character. https://civitai.com/posts/24699512
@silvermoong Yes i know that. As i said, the issues is that this doesn't happen as strongly with other models, as you can see that it still generates extra characters with just (( )). This issue does not happen with other models with the same prompts as easily (be it illustrious, NAI or even other v-pred versions), as they generally only appear after (( )1:3) (hence why I included both). And it doesn't matter what words you use, it happens regardless if you follow the same pattern.
@richerich I see. I guess this is because my model follows prompts precisely or is more sensitive to prompts and weights than other models.