0xSoul 发表于 2023-6-7 22:28:16

[SD 1.4][TextualInversion]Aidra Fox v2.1

<h1>Aidra Fox</h1><h3>UPDATED</h3><p>I re-ran this TI using a smaller pool of source images, getting less wordy in the source prompts, and keeping a close eye on the embedding strength. I think this one looks better, esp with a photorealistic model.</p><p>Training ran for 10,000 steps at a learning rate 0f 0.004, choosing the 6500-step version, based on SD-1.4</p><p></p><h3>(ORIGINAL NOTES)</h3><p>Initial attempt at a TI for the goddess Aidra Fox.</p><p>This is my first ever TI, so let me know how it works out for you.</p><p>I trained her for 12,500 steps but the 4k-7k range looks the best to me. This upload represents the 5000 step version. I used 24 pictures, sfw where possible, for training.</p><p>I used the base v1.5 SD model for training, but <em>highly</em> recommend using any of the top rated checkpoints here.</p><p>PLEASE send along any tips and advice you have, I want to make this model as good and foolproof as possible, thanks!</p>
页: [1]
查看完整版本: [SD 1.4][TextualInversion]Aidra Fox v2.1