接触这东西有段时间了，首先感谢下哔哩哔哩上的秋葉、青龙等等大佬啊,C站上的TASUKU2023（chilloutmix）和 CommuNisM727这家伙人很好还有我师父kim啊这些都求学过，也感谢Q群852450267里的各路神仙的大佬指点!
但是还是我师父的一句话师傅领进门修行在个人哈哈也就是还的靠自己，毕竟没人会手把手教那累死了，那也说明朽木不可雕也，但是没有这些大佬我也不能有机会发布下自己的模型。
可惜学到现在我依旧不会微调大模型，没办法资质不行- -，别笑！其实到现在我也还在纠结训练素材的分辨率是要和训练参数一直好还是更大一些比较好，玄学吗哈哈
这个模型就算是BETA版本了,个人喜欢古风一些也加了一些我喜欢的面容进去，配合其他LORA我测试了几个还将就，现在素材才一千多张我会慢慢更新这个系列，之后到1万张的时候还打算挑战下SDXL版的，许多地方可能不是很完美但是还是有到了我预期的地方，所以也迫不及待的分享出来，希望你别失望！
采样方法(Sampler)： DPM++ SDE Karras| DPM++ 3M SDE Karras
放大算法：4x-UltraSharp
采样迭代步数(Steps)：20-30
高清修复采样次数：+-10
重绘幅度(Denoising):0.45
宽度 512，高度 768
！上半身构图不建议修脸！
Description
FAQ
Comments (4)
我实在是不会English外国友人机翻吧，不过感觉没有外国人会喜欢吧!
感谢分享
Translation from GPT：I’ve been involved with this for a while now. First and foremost, I want to express my gratitude to the greats on Bilibili, such as 秋葉 and 青龙, as well as TASUKU2023 (chilloutmix) and CommuNisM727 on the Civitai website. This guy is really nice, and I’ve also learned a lot from my master, Kim. I’m also grateful for the guidance from all the experts in the QQ group 852450267!
However, it’s still my master’s words that resonate with me: “The master leads you to the door, but the cultivation is up to the individual.” In other words, you still have to rely on yourself. After all, no one will teach you hand in hand, that would be too exhausting. That also means that some wood can’t be carved. But without these greats, I wouldn’t have had the opportunity to release my own model.
Unfortunately, up to now, I still can’t fine-tune large models. There’s no way around it, my aptitude just isn’t up to par. Don’t laugh! Actually, I’m still struggling with whether the resolution of the training material should be consistent with the training parameters or whether it should be a bit larger. Is it metaphysics? Haha.
This model is considered a BETA version. I personally like the ancient style, so I’ve added some facial features that I like. I’ve tested it with other LORAs and it’s passable. Currently, there are only over a thousand images for the material. I will gradually update this series. When it reaches 10,000 images, I plan to challenge the SDXL version. Many places may not be perfect, but it has reached my expectations, so I can’t wait to share it. I hope you won’t be disappointed!
Sampling Method (Sampler): DPM++ SDE Karras | DPM++ 3M SDE Karras
Upscaling Algorithm: 4x-UltraSharp
Sampling Iteration Steps: 20-30
High Definition Restoration Sampling Times: ±10
Redrawing Range (Denoising): 0.45
Width: 512, Height: 768
Please note: It’s not recommended to fix the face for upper body compositions!
特别喜欢，谢谢！
Looks like we don't have an active mirror for this file right now.
CivArchive is a community-maintained index — we catalog mirrors that volunteers upload to HuggingFace, torrents, and other public hosts. Looks like no one has uploaded a copy of this file yet.
Some files do get recovered over time through contributions. If you're looking for this one, feel free to ask in Discord, or help preserve it if you have a copy.
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.