![]() ![]() If the data is highly curated that every single person looks way above average, prompting “woman” would be the same as prompting “beautiful woman”. Open AI trained the CLIP model with proprietary data. This affects the embeddings of the model. The first suspect is switching from Open AI’s CLIP model to OpenCLIP. Filtered out NSFW contents in training data.This is an area I can only speculate… but why not? The two changes in v2 are Why does negative prompt become more important in v2? It is as if v1.5 model does not understand these words. Adding the negative prompt ugly, deformed and disfigured may improve things but it is not as clear as in v2.1. The images comes out pretty good without any negative prompts in v1.5. Later steps are only finer adjustment to details, such as hairs covering ears.Īdding negative prompt to v1.5. The reasoning behind this is that the diffusion process is most important in the beginning steps. Since, I am using 20 sampling steps, what this means is using the as the negative prompt in steps 1 – 10, and (ear:1.9) in steps 11-20. You will get the same image as if you didn’t put anything. You can verify its uselessness by putting it in the negative prompt. Let’s pick the as the meaningless, dud negative prompt. You can use keyword switching to first use a meaningless word as negative prompt, and then switching to (ear:1.9) at a later sampling step. Now what if you really want to use a high emphasis (ear:1.9)? I don’t know what’s your problem with ears but I have a trick for you. Negative prompt could affect the diffusion process strongly. The ears are covered more by hair with in all emphasis factors but when the factor reaches 1.9, the composition of the image changed. Below are with three increasing emphasis 1.3, 1.6 and 1.9. What if we are ok with the wind but want the hair to cover the ear? Let’s add negative prompt “ear” with different emphasis factors.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |