Stable Diffusion 2 Prompt Convertor

Key Takeaways

hey friends welcome to one little coder in this table diffusion tutorial we're going to see how to create better prompts for stable diffusion 2.0 one of the complaints a lot of people have been talking about for stable diffusion 2.0 is that the previous prompts that they were using they had mastered using stable division 1.4 or 1.5 do not work well for stable division 2.0 another the primary reason for that is the text encoder which actually converts takes your prompt maps to the image it's called text encoder so this text encoder previously stable division was using one by open AI the clip encoder but now they have moved on to open clip which is a different text encoder which means it's the text that you are using as prompt for your images are not going to work and that is the first problem and then the second problem is the new stable division training data has eliminated couple of artists which means for example if you are using grid root cause 3 previously that artist name is not going to be usable anymore in the new model in itself so all these problems have created chaos and confusion in the community that you cannot use existing prompts thanks to this new tool that is built by silver and filoni who uses clip interrogator and tries to convert the 1.5 or 1.4 prompt to a prompt that is compatible with 3.0 first let us see a quick demo of how it works so you've got a text box here I'll link this hugging face in the YouTube description please check it out so you can go here and enter your stable division 1.4 or 1.5 prompt text and then click submit and then it is going to convert that into it so the way you can use it is you can go to lexicon.art where you've got a lot of prompts copy The Prompt that you like and then come back to prompt converter and paste your prompt here and then click submit once you click submit it is going to create the new prompt that is compatible with stable definition 2.0 now copy this prompt go to your stable diffusion and then paste this prompt and then see how it is going to work I have like found out that certain prompts really work fine um in my experiment that I've tried but again given that stable division requires you to have a negative from or a negative prompt is quite essential for stable division 2.0 then stable division 1.5 I think you need to add stable you need to add negative forms for you to get better image but you can actually see certain problems for example if an artist's name is not there that is solved if if something that was not very compatible with stable division 2.0 that is also now you get a better prompt now all you have to do is change negative prompt and then get a better image in itself so how does this work the way this works is it takes this prompt creates an image using stable division 1.5 and then sends that image to clip interrogator which is another very good application and that application takes this image and then reads this image and tries to understand it and then creates a prompt based on it and that prompt is written to you actually like you have a prompt that is for 1.5 or 1.4 now a stable division 1.4 1.5 model is used to create a new image that image is fed into the clip in derogator which is going to read the image and give you a open clip compatible prompt and that prompt is now given back to you here which you can use to create a beautiful stable division 2.0 see this is not a foolproof solution that is going to make your prompt amazing but this is going to solve certain problems like the kind of language that you use the art is that you are missing it's going to solve that problem but you still have to go over the things like negative problems for example I can say negative from 3D ugly and then I can again try to generate it so that is going to create a better image and using the prompt as it is but this actually solves the problem that everybody has been talking about that your existing prompts do not work very well and this is a wonderful opportunity for you or Forest as a community to improve ourselves to create better um better prompts and you can as you can see like while we are talking this image looks really beautiful to me I don't know whether it is exactly what that the original prompt could have created but I am actually um honestly happy with the kind kind of image that it has generated here and I would love to know if you have found this useful or if you have any other prompt tricks that you would like to share it with the community um this link and everything else will be in the YouTube description please check it out otherwise see you in the next video peace