Hello everyone, I'm Moogone and I'm back again. Today I bring you a big model that was generated using over 1500 images with blip text. It can be described as a large collection of exhibition halls/museums, and you don't even need to trigger it with any keywords. You can simply use natural language to generate the desired results.
However, please note a few usage details:
1. Choose "Euler a" as the sampler.
2. Check the "Hires. fix" option.
3. Select "None" as the "Upscaler" in Hires. fix option. (This is very important, otherwise overfitting may occur)
Please post the generated results with me and honestly rate them. Thank you very much. Your ratings and sharing of results are the driving force for me to continue developing the next generation of larger sample models. The next sharing will be after the number of raters exceeds 10, and at least 5 people post their images. If you have any questions, you can add my WeChat: Moogone.
大家好,我,Moogone又回来了,今天我带来的是一个超级大模型,这个模型用了1500多张带文本的图像生成,可以称之为展览馆/博物馆大集合,你甚至不需要用任何关键词触发,你只需要用自然语言就能生成想要的结果。 但是要注意一些使用上的小细节:
1. 采样方法选择 “Euler a“
2. 打开√“高分辨率修复”选项
3. 高分辨率修复的放大算法选“无”(这一点很重要,不然会出现过拟合)。
这次拜托大家将生成的结果发上来,并且诚实地给我打分,谢谢大家了,你们的打分及成果分享是我继续研发下一代更大样本模型的动力,下一次分享将在评分人数超过10人,以及超过5人分享图片后。 如果遇到任何问题可以加我的微信:Moogone,合作及定制企业模型可以联系我。