From Sketch to Art: Stable Diffusion and ControlNet Magic Tutorial

ever wondered how to breathe life into your sketches well today we're about to reveal the magic using stable diffusion and control net stay tuned I will create a new document in Photoshop with a size of 1024 pixels then I will create a new layer with vertical symmetry activated and use a pencil looking brush to start sketching my ideas I prefer sketching in Photoshop because it allows me to utilize symmetry draw and arase quickly warp cut and paste parts to adjust the sketch according to my preferences I use a waycom tablet connected to my computer for this process I like to begin by sketching some larger shapes to establish the proportions and composition after that I gradually add more details the entire sketching process usually takes me around 10 minutes because I enjoy experimenting and may not have a clear idea of what I want initially I simply follow the flow and play around with shapes until I see something I like now let's save the image as a JPG or PNG and proceed to The Next Step I'm using automatic 1111 which is installed on Windows for the checkpoint I'm using an sdxl model called Juggernaut XL which you can download from the civit AI website it is currently on version 6 and the size on dis is approximately 6 GB to use it you will need to place it in your stable diffusion folder by navigating to the web UI folder then models and and finally stable diffusion I'm also utilizing two additional extensions in this video one of them is called sdxl Styles and you can find a tutorial video about it on my pixaroma channel titled installing style selector sdxl this extension enables me to quickly experiment with different Artistic Styles the most crucial extension for this process is control net I've already provided instructions on how to install it in my previous video transforming sketches into vibrant artart I asked chat GPT to provide me with some prompt ideas based on the description I gave it in my description I included details on how I wanted my final illustration to look first I'll copy the initial prompt and see what results we obtain I'm using a text to image method and I paste The Prompt I'll also specify the settings for sampling method steps and size uh 1024 pixels for the artistic style I'll choose a fantasy art style uh moving on to control net I'll upload my sketch and check the enable and Pixel Perfect options I'll select canny as the control type and the pre-processor will also be canny as for the model I'll use Coya control light XL after setting these parameters I'll hit the generate button and within seconds we'll have our first results while the initial illustration looks good I'll click generate a few more times to see how it varies with different seeds when I find one I like I'll aim for a larger version of it to do this I'll use the same seed and activate highers fix for the model I'll switch to 4X Ultra sharp and set the Deno strength to 0.45 then I'll generate a larger image with a resolution of 248 pixels now let's try a different style such as cinematic which adds a more photorealistic touch I'll use a random seed again and deactivate Iris fix you can see how it starts to resemble a movie scene I'll experiment with a few more prompts different seeds will prod produce diverse results so if you're unsure about what you want you can keep generating until you find something you like you can also experiment with negative prompts assign weights to words that are crucial to your vision or edit the prompt to guide the output towards your desired outcome in this video my primary goal was to explore a variety of creative ideas so here's another illustration that I like first uh I reuse the same seed then I activate uh ties fix and hit the generate button uh to double the size uh you'll notice it now has much more detail and is sharper I enjoy experimenting with various prompts and art styles you can adjust the prompts yourself or ask chat GPT to help you refine them it's kind of fun to explore and see what you can create keep in mind that the AI Works around your initial sketch and ideas it brings your Concepts to life but the creative process isn't entirely controlled by the AI you can guide it to achieve your vision once you have an illustration ready you can further refine it in Photoshop and then bring it back to image to image to get variations of it here's where I switch to image to image to create some variations the great part is that you can change the style for example if you have a cinematic photo like in this case you can transform it into a fantasy illustration using image to image play with the denoy strength and use canny but this time with the actual color illustration the hyperrealism style also yields impressive results and techware fashion produces some captivating futuristic images I could keep experimenting with this all day but let's explore one more style this time I'll ask chat GPT to help adjust the prompt to create a cute and cartoonish illustration for the style I'll opt for the digital art style style which tends to work well with various types of illustrations I'm pleased with the second generation I received so I'll use higher fix on this one as well to obtain a larger sized image before I conclude this recording I'd like to share one more technique you can use to fix minor errors in Photoshop I'm particularly fond of the remove tool which is available in the latest versions of Photoshop it's significantly more effective than the spot healing tool and allows for precise cleanup of the illustration alternatively depending on your skill level you can also employ digital painting to cover any imperfections if you enjoyed watching this video and found these techniques inspiring don't forget to show your support by liking and subscribing to the pixaroma channel by subscribing you'll stay updated and won't miss out on uh more exciting content like this in the future thank you for joining me on this creative journey and I look forward to sharing more with you soon

Channel: pixaroma

Views: 11,719

Keywords: stable diffusion, stable diffusion automatic1111, stable diffusion automatic1111 tutorial, stable diffusion tutorial, stable diffusion video tutorial, from sketch to digital art, from sketch to ai art, from sketch to art, ai art, stable diffusion ai, sketch to render ai, stable diffusion ai art, fantasy art, control net, controlnet tutorial, canny controlnet, sdxl controlnet canny, control net tutorial, automatic1111 tutorial, pixaroma, ai image generation tutorial

Id: IBNuALJuOgw

Length: 7min 9sec (429 seconds)

Published: Mon Oct 30 2023