santa hat
deerdeer nosedeer glow
Sign In

TRIGGER ONLY Character lora:Jinx Default | League of Legends仅触发词角色lora:金克斯默认 |英雄联盟

TRIGGER ONLY Character lora:Jinx Default | League of Legends仅触发词角色lora:金克斯默认 |英雄联盟

Jinx Default | League of Legends

Total score: 2.3 (3 stars) available

Details: 3 available

Most of the images are on the younger side and have darker skin color. Weapons are drawn in Mechamix, but they are changed a lot. The characters have obvious expressions on their faces, but there are some differences from Jinx.

Style: 2 partially available

You can get 2D images, 3D images are close to game modeling, and there is no live-action photography style.

Failure rate: 2 partially available

There is a clear phenomenon of multiple heads. There is a missing lower body in the picture. Curious how the author handles the training set?

The test used only one trigger word, generating 40 images with 4 seeds and 10 models. Examine the initial performance of the character LoRa itself. There are three objective dimensions: detail, style, and failure. 1 Not available 2 Partially available 3 Available 4 Excellent 5 Perfect . The overall score matches Civitai's five-star rating.

总分:2.3(3星 )可用

细节:3 可用

大部分图像的年龄都偏小,并且皮肤颜色较暗。在mechamix中画出了武器,但被改变较多。人物面部有明显的表情,但是和Jinx有些区别。

风格:2 部分可用

能得到2D图像,3D图像接近游戏建模,没有真人摄影风格。

失败率:2 部分可用

有明显的多个人头现象。有画面内的下半身缺失。很好奇作者是如何处理训练集的?

该测试仅使用单个trigger word,用4个种子和10个模型生成了40个图像。考察人物LoRa 自身的初始性能.有三个客观维度:细节、风格和失败。1 不可用 2 部分可用 3 可用 4 出色 5 完美 。总分和Civitai的五星匹配。

2

Comments