这个模型生成的结果有些糟糕，可能需要您手动地对分割后的图像进行排序以达到较好的效果。
这个Lora模型是基于final-prune.ckpt（【旧短哈希1d4a34af】新短哈希22fa233c2d）训练出的。
1.使用768x768的分辨率生成图像
2.使用photosop等软件分割图像
3.分别保存这些图像并使用诸如ezgif的网站或者工具来创建你的 gif 动画
您也可以使用1024x1024的分辨率生成图像，但这可能会生成不稳定的结果。
灵感来源自easily animated deepthroat concept style [TEST].
推荐权重为0.8。
英语是用Google翻译翻译的，如果有造成误解之处非常抱歉。
This model produces somewhat poor results and may require you to manually sort the segmented images to achieve good results.
This Lora model is trained based on final-prune.ckpt ([old short hash 1d4a34af] new short hash 22fa233c2d).
1. Generate images with a resolution of 768x768
2. Use software such as photosop to segment the image
3. Save these images separately and use a website or tool like ezgif to create your animated gif
You can also generate images with a resolution of 1024x1024, but this may produce unstable results.
Inspired by easily animated deepthroat concept style [TEST].
The recommended weight is 0.8.
English is translated by Google Translate, sorry if there is any misunderstanding.
Description
FAQ
Comments (9)
How did you train it? I wanna try to train a grid Lora myself
English is translated by Google Translate, sorry if there is any misunderstanding.
First, prepare videos or gifs of concepts that meet your requirements, split them into frames, and then use the batch cropping tool to crop the resulting frames into square images.
Then use the collage tool to stitch them into 3x3 (768x768) or 4x4 (1024x1024) grid images, and the data set is processed, just train it like other lora.
I prepared about 50 images for this lora training, the concept folder is 6_, trained 16 epoches, and used the lion optimizer. Finally, I picked out a file that was saved every 2 steps with a relatively good effect. out.
英语是用Google翻译翻译的，如果有造成误解之处非常抱歉。
首先，准备好符合你的要求的概念的视频或者gif，把它们分割成帧，然后用批量裁剪工具把得到的帧裁剪成正方形的图像。
接着使用拼图工具把它们拼接成3x3（768x768）或者4x4（1024x1024）的宫格的图像，数据集就处理完毕了，像训练其它lora一样训练就行了。
我这个lora训练时准备了约50张图像，concept文件夹标准了6_，训练了16个epoches，使用的是lion优化器，最后从每2步保存一次的文件中挑出了一个效果相对较好的出来。
For anyone who needs a way to combine the images into a gif, use ezgif's sprite-cutter. Works online and with good results.
请问能分享一下训练这种类型的方法吗
首先，准备好符合你的要求的概念的视频或者gif，把它们分割成帧，然后用批量裁剪工具把得到的帧裁剪成正方形的图像。
接着使用拼图工具把它们拼接成3x3（768x768）或者4x4（1024x1024）的宫格的图像，数据集就处理完毕了，像训练其它lora一样训练就行了。
我这个lora训练时准备了约50张图像，concept文件夹标准了6_，训练了16个epoches，使用的是lion优化器，最后从每2步保存一次的文件中挑出了一个效果相对较好的出来。
Here is a simple Python script to convert a PNG image (with any number of tiles) to a GIF.
The parameters are: H tiles, W tiles, image name, FPS . Example: python GenGif.py 3 3 input.png 4
实际操作还是很难的
脸崩的厉害得修，出来的姿势不连贯得凹，而且看起来流畅帧数还不能低了
Just use gif to gif in A1111.
太6了，正好想做这个
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.