Additional training and merged models based on the SD1.5 model are available.
The model merge has many costs besides electricity. With your support, we can continue to develop them.
No baked VAE.
The recommended VAE is "vae-ft-mse-840000-ema-pruned.ckpt".
If you try it and make a good one, I would be happy to have it uploaded here!
The merging source is listed in each version detail.
モデルマージは、電気代以外にも多くのコストがかかります。皆さんの応援があれば、これからも開発を続けることができます。
https://ko-fi.com/breakd
VAEは焼いてません。
おすすめのVAEは"vae-ft-mse-840000-ema-pruned.ckpt"です。
使ってみて、いいのができたら、ぜひここにアップしてもらえると嬉しいです！
Description
・_BreakTheDarknessV07をマージ
FAQ
Comments (1)
Hi there!
So sorry to bother you, I am having several issues with your model and the configuration of it all. I have tried (inside of comfyUI) the following samplers and schedulers:
Euler: normal, Karras, exponential, sgm_uniform
DPM++ 2M: normal, Karras, exponential, sgm_uniform
Along with the VAE: ft-56000; ft84000; kf-tl8 anime vae; orangemix.vae
So far, I have failed to generate any kind of results or good images with Euler, which I attribute to the sampler the model used during training, but I have also struggled with 2M. Some times during generation the model produces a good batch of four images (usually under 56000 vae or kf anime vae or orangemix vae) but it misses, mouths, nose and even basic face details that I prompt for
On the other hand, when using this model with 840000 I do get more colors, but the face distorts entirely, no eyes, no mouth, or tiny eyes with no other facial structures.
While I am not entirely as to why this happens, I wanted to politely ask for your input and information regarding best sampler, scheduler, vae and negatives or positive embeddigs you might use.
Since I am struggling to replicate the examples shown here and on the community category of the model.
Again, thanks in advance.
Volv1t3