好久不见，甚是想念
BEYOND REALITY系列模型断更已经半年多，主流开源绘图模型也已经快速迭代到了Z时代。
Z IMAGE给大家带来了足够的惊喜，超快的速度和超真实的画面，社区中对edit的呼声更是持续不断。
我知道你们很着急，但是先别急，先来看看社区微调版的 Z image 哈哈哈
BEYOND REALITY Z IMAGE 1.0 浓妆淡抹总相宜，这是一个基于人像数据集微调lora合并而来的模型，延续BEYOND REALITY系列模型的特点，高清晰、高美学。Z IMAGE目前的问题之一就是高频信息的表现不佳，体现在画面上就是纹理的细节不好。因此训练了一个lora来提升模型的纹理质量。本想发布lora的，后来想想，一直都是发布checkpoint的，那还是延续这个习惯吧。
我一直偏爱胶片摄影的光影色彩，因为那是无数美学工程师用百年的调色经验沉淀下来的美学的体现，是用图片记录光影色彩、记录美的最佳的、受众最广泛的体现。美不是被定义的，而是被感受的，能被定义的或者量化的，一定是会沦为平庸。所以美无法量化无法评估，只能去理解和感悟。美也无法让所有人都共情，能遇到一个人理解你，已经是万幸之事。
本次更新主要优化Z的人像类皮肤纹理和环境纹理细节，并优化模型生成人像类图像的美学。Z 本身生成的图片真实度非常高，某些情况下会让你完全分不清是AIGC图还是真实照片。但是，在美学方面，Z 还有优化提升的空间。
本次更新的模型整体风格略偏胶片人像摄影，可以生成很美的人像类图片，也可以生成很优美的环境细节，同时保留了 Z 的部分真实感。在提示词引导下可以生成非常真实的图片。
注意，因本次训练以合成的高质量人像摄影风格图片为主，在提升整体美学与纹理细节的同时，有可能会对复杂肢体动作和某些非人像类纹理造成影响。后续版本会持续优化。
欢迎体验试用，欢迎多多返图。
BEYOND REALITY，I'm back。
__________________________________________________________________________
Long time no see. I miss you very much
The BEYOND REALITY series models have been out of update for more than half a year, and the mainstream open-source drawing models have also rapidly iterated to Generation Z.
Z IMAGE has brought everyone a lot of surprises, with its super-fast speed and ultra-realistic visuals. The calls for edit in the community have been continuous.
I know you're in a hurry, but don't worry yet. Let's take a look at the community's fine-tuned Z image first. Hahaha
BEYOND REALITY Z IMAGE 1.0 looks good with both heavy and light makeup. This is a model that is fine-tuned based on the portrait dataset and merged with lora. It continues the characteristics of the BEYOND REALITY series models, featuring high clarity and high aesthetics. One of the current problems of Z IMAGE is its poor performance of high-frequency information, which is reflected in the picture as poor texture details. Therefore, a lora was trained to improve the texture quality of the model. I originally intended to release lora, but then I thought that since I've always been releasing checkpoint, I might as well continue this habit.
I have always been fond of the light, shadow and color of film photography, because it is the embodiment of aesthetics accumulated by countless aesthetic engineers with a century of color-grading experience. It is the best and most widely accepted way to record light, shadow and color, as well as beauty through pictures. Beauty is not defined but felt. What can be defined or quantified is bound to become mediocre. So beauty cannot be quantified or evaluated; it can only be understood and perceived. Beauty cannot resonate with everyone. It is already a great fortune to meet someone who understands you.
This update mainly optimizes the skin texture and environmental texture details of Z's portrait type, and improves the aesthetics of the portrait type images generated by the model. The images generated by Z itself have a very high degree of authenticity. In some cases, you can't tell at all whether they are AIGC images or real photos. However, in terms of aesthetics, Z still has room for improvement and refinement.
The overall style of the model updated this time is slightly more like film portrait photography. It can generate very beautiful portrait pictures as well as very elegant environmental details, while retaining some of the realism of Z. Very realistic pictures can be generated under the guidance of prompt words.
Note that as this training mainly focuses on high-quality synthetic portrait photography style images, while enhancing the overall aesthetics and texture details, it may also have an impact on complex body movements and certain non-human portrait textures. Subsequent versions will continue to be optimized.
Welcome to try it out and send back more pictures.
BEYOND REALITY, I'm back.
