<h2 id="heading-241">使用须知：权重使用0.8最佳，底模仅推荐使用pony系列的模型，T-ponynai3的模型最佳（毕竟是训练底模）</h2><h2 id="heading-234">Instructions for use: A weight of 0.8 is the best, and only Pony series models are recommended for the base model. The T-ponynai3 model is the best (after all, it is a training base model)</h2>为了能更好的在pony上使用不同的画风，我选择通过训练额外的风格lora的来达到这一目的(显卡性能不够训练更多的训练集花费的时间成本过高了），同时为了避免侵害到画师的版权，我将会把每个不同的画风lora命名为style_number，这个合集我将会持续更新，训练lora所需的时间相较于训练底模可短多了。In order to better use different art styles on Pony, I have chosen to achieve this goal by training additional styles of Lora(The graphics card's performance is insufficient, and the time and cost of training more training sets are too high.). At the same time, to avoid infringing on the copyright of the artist, I will name each different art style Lora stylenumber. This collection will continue to be updated, and the time required to train Lora is much shorter compared to training the base model.

T-ponynai3-style-lora

触发词style_2可用可不用，会略微影响效果强度，推荐权重0.8左右The trigger word style_2 can be used or not, as it may slightly affect the strength of the effect. The recommended weight is around 0.8

tpony-style-v2.safetensors

style_2

style_4

style_3

style_1

使用须知：权重使用0.8最佳，底模仅推荐使用pony系列的模型，T-ponynai3的模型最佳（毕竟是训练底模）为了能更好的在pony上使用不同的画风，我选择通过训练额外的风格lora的来达到这一目的(显卡性能不够训练更多的训练集花费的时间成本过高了），同时为了避免侵害到画师的版权，我将会把每个不同的画风lora命名为style_number，这个合集我将会持续更新，训练lora所需的时间相较于训练底模可短多了。

 T-ponynai3-style-lora

触发词style_2,可用可不用，用了效果更强，权重0.8

Files with SHA256: 43a55b5cd652392d4deece372baf434ec9e116e7e5d5137c6417095a7da86a7e