What is bad_prompt_version2 and how to use it?

5/5 - (1 vote)

bad_prompt_version2 is a negative embedding for text-to-image diffusion models, such as Stable Diffusion. It is trained on a dataset of images with undesirable artifacts, such as malformed hands, distorted faces, and strange compositions.

The purpose of bad_prompt_version2 is to help the model generate images that are more realistic and aesthetically pleasing.

To use bad_prompt_version2, simply add it to the negative prompt of your text-to-image diffusion model. For example, if you are using Stable Diffusion, you would add the following to your prompt:

(bad_prompt_version2:0.8)

This will tell the model to avoid generating images that contain the artifacts that are present in the bad_prompt_version2 dataset.

How to use bad_prompt_version2:

If you wan to use the bad_prompt_version2 then you can follow these steps to get the best result. I have provided the examples of the bad_prompt_version2 so you can use it accordingly. Here are some examples of how to use bad_prompt_version2:

  • To generate a portrait of a person with realistic hands, you could use the following bad_prompt_version2 prompt:
a portrait of a person with realistic hands (bad_prompt_version2:0.8)
  • To generate a landscape image with realistic clouds, you could use the following bad_prompt_version2 prompt:
a landscape image with realistic clouds (bad_prompt_version2:0.8)
  • To generate a product photo with a clean and professional look, you could use the following bad_prompt_version2 prompt:
a product photo of a [product name] with a clean and professional look (bad_prompt_version2:0.8)

Tips to use the bad_prompt_version2:

  • Start with a strength of 0.8 and adjust it as needed. A higher strength will make the model more likely to avoid the artifacts in the bad_prompt_version2 dataset, but it may also make it more difficult to generate the desired image.
Also Check  Complete guide to sell your prompt Online 2023

What to do if I don’t getting the result I want.

  • If you are not getting the results you want, try adding other negative prompts to your query. For example, if you are trying to generate an image of a person with realistic hands, you could add the following negative prompts:
(malformed hands:0.8)
(deformed hands:0.8)
  • You can also use bad_prompt_version2 in conjunction with positive prompts. For example, if you are trying to generate an image of a black cat, you could use the following prompt:
a black cat (bad_prompt_version2:0.8)

This will tell the model to generate an image of a black cat, while avoiding the artifacts in the bad_prompt_version2 dataset.

Conclusion

bad_prompt_version2 is a powerful tool that can help you generate more realistic and aesthetically pleasing images with text-to-image diffusion models. By using bad_prompt_version2 in your prompts, you can avoid generating images with undesirable artifacts, such as malformed hands, distorted faces, and strange compositions.