Verified: a month ago
SafeTensor
The FLUX.1 [dev] Model is licensed by Black Forest Labs. Inc. under the FLUX.1 [dev] Non-Commercial License. Copyright Black Forest Labs. Inc.
IN NO EVENT SHALL BLACK FOREST LABS, INC. BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH USE OF THIS MODEL.
去除原图中大部分的光影、渐变，将输入的图片转换为低饱和度的平涂线稿风格。
It removes most of the lighting and gradients from the original image, converting the input into a low-saturation, flat color lineart style.
初衷并不是为了作为风格lora来训练的，我在通过kontext使用lora来进行画风转换时遇到问题，如果原图和目标画风的人物整体差异不大，kontext似乎更倾向于仅修改面部，却保留身体和服装的所有细节，导致脸部与身体有很明显的割裂感。
The initial intention was not to train this as a style LoRA. I encountered a problem when using LoRAs for style transfer via kontext: if the overall difference between the character in the original image and the target style is minimal, kontext tends to modify only the face while retaining all the details on the body and clothing. This creates a very noticeable, disjointed look between the face and body.
通过先提取线稿再进行画风转换可以一定程度上解决这个问题，但是会丢失所有颜色信息，所以尝试训练了这个lora，可以尽可能只保留原图的线条和颜色，来提高其他风格lora的效果。
While extracting the lineart first before the style transfer can partially solve this problem, it results in the loss of all color information. Therefore, I tried training this LoRA, which can preserve the original image's lines and colors as much as possible to improve the effectiveness of other style LoRAs.
使用Civitai上的两个lora进行对比测试：
Using two LoRAs from Civitai for a comparison test:
