Unlock the Power of AI Image Generation with Griptape and ComfyUI: Create Powerful Text Prompts

hello there I am super happy to show you four different workflows using grip tapes framework in conf UI to create amazing very powerful and customizable tax prompts for image generation let's check it out the first one is a basic image description the second one how we can create more details the third one how we can actually extract some elements and the fourth one how we can combine image descriptions from different images into one let's build a workflow really quick you can double click and go UI and type image description grab the node and connect the image into the input image of the image description double click and type agent make sure it's a create grip tape agent you want to graph the output in to the input of the agent double click and create rules now route the rules node and connect it into the rule set of the agent itself last but not least double click and type text go down and search for grip tape display text and you want to Output into the input now we want to give some rules to the agent here I'm going to say you are an expert at creating amazing cinematic text prompt for film like high quality images add as much detail as possible then on the agent itself I'm asking it to summarize the entire text and give it an example of how you can actually create a good prompt after that we can run and see what we get and there you have it that's our new beautiful prompt the last part is that you want to connect the prompt that you just generated in into one of your workflows in this case I have a basic image generation one for that we need an extra node double click type click Text and code from grip tape it's very straightforward output to input and now the output of the text end code goes into the positive prompt of the K sampler and also the clip from the checkpoint model and there you have it now we can run now let's go back to the other workflows to see what else we can do perfect let's see what I did with the first one we can see we have just like the example I just describe a image description then rule set the grip T agent it creates a very interesting prompt fitted into image generation as you can see there are differences between between those two images let's try something else to make it look more like the original image now instead of having only one image description I have three so one two and three then I am using this amazing merge text Noe that will allow me to combine all those image description ones in that agent on the first description I have described the pH in detail the second one says describe the camera angle and the third one it says describe the background in detail so after that with the same agent with the same rules I get a little bit more extended uh image description then I fit it into the same image generation and look at that I have a little bit more similarities between the new one and the original image on the third one I wanted to do something a little bit different I wanted to extract the background only so in the first image description you can see describe only the environment not people and the second one it says describe only the time of day we run with the same agent the same rules we get a different prompt and there you have it the image it's only the background it's not identical but you know there's definitely room there to play around with the settings and the rule sets and and things like that what if you wanted to try something even a little bit more complex now I have three different images for this first image description I am asking it to describe only the car the second one is just the background and the third one it's the lighting and mood color after that the same approach merge the text into the prompt I'm using the same rule set the same agent it creates a highly image of a single suer but then describes the background and then describes the lighting which is amazing then we fit it into the same image generation workflow after that you can upscale it and generate something like this remember this is just one way to use it but there are so many different workflows inside of comfy UI we can't wait to see you try and create amazing things than thank you

Channel: Griptape

Views: 5,851

Keywords: Griptape, ComfyUI

Id: BoOlDSGivms

Length: 5min 43sec (343 seconds)

Published: Thu Jun 06 2024