Memo-Ray is a anime model finetuned on NAI.Carefully selected many characters and illustration images for training purposes.Improved certain issues present in traditional anime models.Compared to the Merge model, the Merge model tends to be overfitted. It does not always adhere to your prompts and may add irrelevant details. In contrast, the finetune model does not have these issues.In addition,don't download V1 and V2. It.s trained On 768 resolution(failure).<h2 id="heading-607">Feature:</h2><ul><li>Trained on 1024 resolution.Useing higher base resolution to generate images.</li><li>Offset noise.</li><li>Better background And scenery</li><li>Better time concept. addresses the limitations in accurately generating night, twilight.</li></ul>Of course, there are many other aspects of this model that I haven't specified. However, given the limitations imposed by Civitai's guidelines, I cannot provide further details here. Including NSFW content. I will not provide further details on this matter.You are encouraged to explore and discover them on your own.<h1 id="heading-114">NOTE:</h1>I strongly recommend using hire.fix to generate images.Because of offset noise include the model,so it is not compatible with LoRA that also include offset noise.An evident manifestation would be color distortion and blurred images. If you insist on using both, the suggested solution is to download the <a target="_blank" rel="ugc" href="https://civarchive.com/models/13941/epinoiseoffset">offset noise lora </a>and set the weight to a negative value, such as -1 or -2.And refrain from using this Lora when using the model under normal circumstances.If you wish to merge this model and release, please mark out the model.

Memo-Ray

Optimized hyperparameters and expanded the training dataset, then retrained the model.

Files with SHA256: b0a8d6250106e41c7ca64eb080f0b1dc4b16063022c1e700db680029e20939e3