Stable Diffusion IMG2IMG HACKS You need to TRY!

these artworks were generated with long bizarre and incomprehensible prompts yes they're definitely awesome but there has to be an easier way to get amazing results with stable diffusion without having to spend hours on Crafting long anime late prompts that's why I developed revolutionary simple new way to add amazing details to your Generations it'll even let you control lighting and composition in your art and I call it Vibe Fusion Vibe ution cner Vibe okay it's really just fing around with image to image and control net but it's really cool and I think you'll like it if you watch my image to image video you're already familiar with this tab in automatic 11 left just drop in an image tweak your prompt and keep iterating until you get something you like cool but our results are a little flat and I want more of a stylized theme to my image I can try to describe the theme in the prompt but it's hard to get the color consistency and vibe that I'm looking for what if there was a way to inject The Vibes that I want right into stable diffusions veins okay head over to your favorite free stock website I'm using pixels.com because it's amazing I want an Autumn theme to this picture with all the Deep Brown and orange colors of that season so let's search for Autumn and see what comes up this looks good let's take that and drag it into the image damage box I'm going to keep this pretty simple base prompt and these settings and hit generate how cool is that now I know what you're thinking how well remember image to image will take color information and composition elements from the image we provide and merge it with our prompt basically image to image has taken the colors from the stock photo and Infused it with our words it's even affected the composition of the image by placing the subject in the darker areas and you can play around with this theme as much as you want you can keep your prompt neutral and just drag and drop any image that you like the color scheme this technique Works particularly well with images of texture fouet lights and Landscapes try them all out and see what works it even works with crappy photos you take yourself like this one if you want want you can try to match the image thematically with your prompt to enhance the look while pairing random unrelated images can yield some mind-blowing results you can also create some random crap from Photoshop and see how that works I like using gradients using some color theory rules to see how that can impact the feel of the image you can go even further and sketch out your composition in Photoshop for image to image to use we can tell staple diffusion where we want the subject to be by using skin tones and we can use different colors for different elements like clothes this is basically like using the sketch t an automatic 1111 but we can get better and more accurate results with tools in Photoshop that's cool but these images are all random what if I want to keep the elements of a picture I like well disembodied AI voice that's where control net comes [Music] in the control net extension and automatic 1111 lets us leverage the control Net Technology to get better control over our images to use the control net extension you need to install it from the extensions tab click the tab then available then load from search for sd- web ui- control net and hit install you'll need to download models that are specific to control net for this to work head over to the hugging face link in the description to find the files you need you can download them all but for this tutorial we just need the depth files once they're downloaded open your automatic 1111 folder then extensions then open the control net folder then models and drop the files in there head back to the installed tab hit check for for updates and then apply and restart UI once you've restarted you'll find the new control net panel under your image to image settings and don't forget to click upload independent control image now control net is huge and we're not going to cover it all in this video let me know if that is something you want to see in the comments but for now I'm just going to show you the settings that work best for this technique put your control image in here I find depth works best for this so we're going to choose that under control type next we can choose a pre-processor we're going to choose minus depth because it seems to work best and it sounds cool make sure to choose the corresponding model the defaults here work fine but play around with the rest of the settings to see what works best for you remember to take enable so control net actually works now drop your stock photo into the image to image box to get a really big impact we'll want to set the denoiser strength relatively High we're going to start with 75 but just know the higher we go the more stable diffusion will take the source image into account for the new generation and now you've infused your new image with the vibe from the old image again rinse and repeat with other stock photos and stuff from Photoshop to get amazing new results there's one other thing we can do with this technique that's super powerful and that's control Lighting in our images open up Photoshop and drag in your control image create a new layer set the opacity on the new layer to round 75% so we can see what we're doing choose the paintbrush tool with theof soft round brush choose black and white in your color palette let's say we want to make some serious contrast with Shadows and Light in our image on the new layer paint in any areas you want to be in Shadow with black and any areas you want to highlight with white I want some lighting Rays shooting in from the side here so I'm going to draw them in like this once you have something you like hide the background layer by clicking the eyeball beside it then click file export and quick export as PNG to save your file now drag that into the image to image box with control net enabled and hit generate play around with the Den noiser strength until you get something that looks good I find anything between s and 9 can work well here you can control the intensity of the light using variations of white gray and black and you don't have to stop with black and white you can use color to get different light effects like everyone's favorite cinematic teal and orange look and that looks amazing and we didn't have to write a huge complicated prompt and it's something that even a beginner can do if you are new to stable diffusion check out this video on how to get started thanks for watching you forgot to ask for likes and subscribes because you're a small insignificant YouTube channel with Big Dreams and you can't do it without support well I wouldn't say insignificant I would if you found this video helpful leave a like And subscribe for more content like this thanks for watching all the way to the end catch you next time

Channel: Incite AI

Views: 23,005

Keywords: stable diffusion, stable diffusion tutorial, ai art, img2img stable diffusion, automatic1111, automatic1111 stable diffusion, automatic1111 tutorial, stable diffusion ai, controlnet stable diffusion, controlnet ai, stable diffusion ai art, ai art tutorial, ai artificial intelligence, ai art tips and tricks, img2img, controlnet, controlnet automatic1111, controlnet stable diffusion install, stable diffusion art, stable diffusion tips and tricks, img2img tutorial, comfyui img2img

Id: 4t6xxyNS_Qo

Length: 6min 15sec (375 seconds)

Published: Sun Aug 27 2023