Unlocking the Secrets of Leonardo AI

I've been searching high and low for my favorite stable diffusion text to image generator and I think I finally found it Leonardo AI it still has a wait list so not everyone can access the program yet but that can't last forever let's look at what makes this text to image generator special and why everyone is talking about Leonardo Leonardo AI is a stable diffusion based text to image generator but it's not a model you've seen before and it's actually not just one model Leonardo has four incredible stable diffusion models from Leonardo's signature to Leonardo creative each model offers its own set of features and benefits that will take your artwork to the next level but that's not all Leonardo hosts other fine-tuned stable diffusion models mostly designed for gaming some allow you to create custom gaming and image assets others let you make pixel art an isometric designs Let's test drive a few of the models and see what we get log in and click launch app if you're not whitelisted sign up for the early bird access and wait for an invite it's worth it and I didn't have to wait long at all you'll see the available models across the top and examples of created images below like with many other sites like mid-journey and Lexica you can use images by other creators as inspiration you can copy the prompts download the images and even remix there are some stunning images to choose from Leonardo seems focused on gaming toggle over to the game assets and you'll have a lot of examples to work with your tools are in the left hand column Leonardo is a well-integrated program and you'll see these tools show up in various places throughout the program let me take you through a few of the tools let's start with AI image generation there are four main Leonardo models let's start with Leonardo's signature model the model is based on stable diffusion 1.5 it's trained on Square images and is considered a generalist model click generate with this model to open the generation workspace in the prop box enter what you want to see and now you have a choice between using the Leonardo style or none let's see what the Leonardo style has to offer now click generate wow without much prompting at all Leonardo gives a realistic beautiful woman in line with what you'd get in mid-journey even without using the Leonardo style the results are stunning but the Leonardo style does bring a bit of realism to the image Leonardo's signature is the base model but is it your only option the Leonardo diffusion model is more contrasty and can create photorealism and artistic style Wheels using the same prompt the diffusion model creates darker shading and Shadows the effect is more like you'd get in a low-key Studio portrait and notice that the native aspect ratio has changed and now is in portrait orientation the Leonardo select style is trained on stable diffusion 2.1 and can also achieve a high level of realism with this prompt I think I prefer the signature style but I'll have to do some additional experimentation finally Leonardo creative also trained on stable diffusion 2.1 interpreter prompt and adds a bit of creativity Leonardo lets you generate images in stable to Fusion 1.5 and 2.1 but I don't know why you would the Leonardo models are so much more Sleek but add Leonardo's style to the basic stable effusion and you'll start to see its style emerge even if some of the classic stable diffusion imperfections start to show there's no doubt Leonardo makes beautiful images and if you don't like any of the Leonardo models that's okay you have other fine-tuned models to play with there are dozens of models many very specific and more seem to be added each day right now Leonardo is free you get 250 tokens each day to explore how many tokens each generation costs depends on the model you use and how many images you make increasing the step count and image Dimensions also cost more the three images I've been making using the default setting have been using two tokens Leonardo has all the tools you'd want in an image generator you can make up to eight images and you have some aspect ratio options the model is trained for specific dimensions so you may get a warning that changing the aspect ratio may be noticeable when upscaling images you can adjust the guided scale and step count to impact the quality of your image the tiling feature is great for graphic designers looking to make textures and backgrounds simply toggle on tiling and write your prompt as normal you can also import an image and fix a seed I didn't have such good luck with the image damage prompting if the weight was too strong it just replicated the image but with a lot of deformities and the prompt didn't seem very influential in the overall image compare the results of Leonardo's image to image generation with mid-journeys using the same prompt image to image prompting isn't quite there yet with Leonardo but there is a feature that could be a game changer when this is ready you can use an image to direct a pose it's not Flawless right now but I can't see where the Leonardo team is going you can also add negative prompts to Leonardo I've done an entire video on negative prompting and stable effusion creators mostly use negative prompts to fix distortions but so far my test day Leonardo has made beautifully clean images here's an example from the Leonardo Community using common stable diffusion negative prompts the image was created in the RPG model remixing into Leonardo without the negative prompts gave me a very different image but it's still beautiful all the tools I could ask for seem to be here so what is prompt Magic toggle on prompt magic and it uses a different process to create your image here's the original and now with prompt magic the results are stunning either way if you want a bit more help with your image prompts click prompt generation you can enter the basic thing you want to see and Leonardo will give you some prompt suggestions let's use the silver-haired model prompt and see what Leonardo suggests well it suggests some different ways of saying silver hair and some alternate settings this is a great feature and you can generate these prompts directly from here no copy and pasting necessary you can generate up to 25 prompt suggestions at a time and right now Leonardo gives you a thousand prompts per month for free so let your imagination run wild but creating images in Leonardo is just the start click on an image and you'll get all the information about the creation you can see the model used and you can copy the link and download the image you can also see any post-processing that you've done it's all in one place click on the image again and you can see more post-processing tools you can delete unzoom and remove the background and you have a couple of upscaling options up scaling makes the image larger and affects the details in the image regular upscaling gets you to 960 pixels but using the alternate upscaling will increase the image size to 26.88 the alternate upscaler looks experimental it's a way of getting closer to your original image details and I really like the ability to toggle back and forth between the options you can also easily remove the background and zoom out unzooming adds a bit more space around your subject this is a feature we'd love to see in other AI image generators in this case the content used to fill the space isn't completely seamless there's a little body deformation but the idea is solid you get 25 post-processing tokens each day most of the upscaling and post-processing options cost one token you may have to be a bit cheesy about which images you choose to work with but that's not all Leonardo has to offer let's look at the canvas workspace where you can do in-painting and out painting out painting makes your image larger place the generation Square where you want to band overlapping with the image in the prompt box tell the AI what you want Leonardo uses stable diffusion 2.1 or 1.5 for in painting and out painting for in painting use the brush and select the area you want to change in this case I'm going to add a jewel to The Click make sure to download the image before exiting the editor it doesn't appear to save to your gallery and before we wrap up this review I've got one more surprise for you Leonardo doesn't just make images you can train your own stable diffusion models for free upload your own images or choose from the images you or other creators have made in Leonardo each model takes up to 40 images and you can train 10 models a month so you can get pretty specific once the model is trained choose select custom model from your list then choose the model you want to use or remix into this model click View and generate with this model enter prompts just as you normally would and use the keyword tag that you set when you train the model I got got mixed results with my models but the women of mid-journey came out the best for the most part Leonardo AI is easy to use with well integrated tools but there are a few limitations image to image prompting is a problem and so is aspect ratio the Leonardo team warns you that using aspect ratios other than the one the model was trained on can result in deformities so for now looks like the default aspect ratios are the way to go this restricts aspect ratios quite a bit Leonardo is fairly quick the generation time is quicker than mid-journey but slower than some of the other stable diffusion models out there most of the time you're not going to have to wait too long for your images this is the first stable efficient model that I truly think has a shot of competing head-to-head with mid-journey it makes clean images without a lot of AI artifacts with simple prompting but it has more models to choose from than mid-journey and uses the staple diffusion tools share your experience with Leonardo AI in the comments below if this video was helpful remember to like the video and subscribe to the channel this is Jen at making the photo let's make something amazing together foreign [Music]

Channel: Making AI Magic

Views: 53,798

Keywords: Making the Photo, Jenn Mishra, Wits End Photography, Digital art, AI-generated art, AI image, digital photography, AI art, AI photography, AI, AI generated art, AI images, Stable Diffusion, stable diffusion ai, stable diffusion video, stable diffusion examples, stable diffusion tutorial, stable diffusion prompt guide, prompt craft, Stable Diffusion prompts, image prompt, best AI images, Creating stunning AI images, negative prompts, Best AI image generator, Midjourney AI

Id: MQhBEz7HQnU

Length: 10min 5sec (605 seconds)

Published: Mon Mar 20 2023