Have you noticed all the exciting new Generative Art tools popping up every month?
I recently came across Leonardo.ai and was curious to see what all the buzz was about. I've been seeing some great things being said about it on reddit and heard some folks even call it a "mid-journey killer"! But is it really that amazing? Well, I decided to take a deep dive, put it to the test and compare it with the trusty mid-journey. Get ready for a fun exploration of Leonardo.ai's features and find out why everyone is talking about it!
But First, Leonardo At Work
Impressed? Let's Get Started with Leonardo.ai
First you need to sign up for a free account. The sign-up process can be a bit confusing. Usually you have to wait to get access but we found a little workaround for immediate access. I’ll lay it out for you here:
- Go to Leonardo.ai
- To the top right hit “Launch App”
3. First Sign up for early Access
4. You will be taken to another screen
5. Hit back and go back to this screen and then hit “Yes I’m whitelesited”
6. You will be taken to the Login in page, here you can simply log in with Google or Microsoft or you can hit signup
7. Once you’re completed, the first time you will be asked to select your interests
Once you've signed up, you'll be taken to the home page, where you'll find a list of featured models and a community feed showcasing recent creations.
Clean User Interface
The user interface of Leonardo.ai is web-based and easy to navigate. The options available to users include the number of images they want to generate, the image dimensions, the guidance scale, and the tiling. On the right side of the screen, users have the ability to input their prompts, select the model and style they want to use, and include negative prompts if desired.
The interface has a clean and organized design, making navigation simple. The functions and features are presented clearly and do not obstruct the user. It is reminiscent of the layout of Midjourney's account page.
Pre-trained Models and Custom Data Sets
Leonardo.ai provides users with a variety of pre-trained models. These models include photorealistic and artistic styles, vintage photography, magical creatures, and paper art, including "Deliberate 1.1," "Vintage Photography," "Luna," and "Paper Art Style." Our testing shows these are VERY good fine tuned models that give specific but very nice output.
One of the most oustanding features of Leonardo.ai is the ability for users to create their own custom data sets and models by uploading photos. And to do this in a relatively simple and fast way. This allows users to train Leonardo.ai in a specific style and get the results they want.
Training Custom Models In Leonardo AI
Training a model in Leonardo AI is a crucial step in creating your desired outputs. There are several factors to consider to ensure a successful training run.
Number of Images
The number of images you use for training will affect the quality of the outputs. A good range is 8-15 images, but feel free to experiment with different ranges. However, it is not advisable to go below 5 images, and going higher than 30 images may not necessarily improve the quality. The number of images also depends on what you are training, with training an object requiring fewer images compared to training a style.
The characteristics of your dataset, specifically the balance between variation and consistency, will also play a role in the training process.
It is important to have a common theme or pattern between your images for the model to learn from. The elements that are consistent between images are what the model will learn and show up in the outputs.
On the other hand, having a good amount of variation between the images will prevent the model from "overfitting" and getting stuck on certain things that may not be desired in the outputs. The things that vary across your images will be more loosely learned and allow the model to put your trained object in new styles and contexts.
There is no perfect answer to the balance between variation and consistency, and it is something to experiment with. For example, if you are training a model on cute animal characters, the character position, style, and image composition would be the consistent elements, while the animal characters and the clothes they wear would be the varied elements.
Another Unique Feature - AI Canvas
Leonardo.ai features an AI canvas, a powerful, simple and easy to navigate editor that allows you to create and edit AI art images in a new and innovative ways. If you have been previously intimidated by other Stable Diffusion models, this could be the solution for you.
You can either upload an image from your computer, a previous generation, or from the community. Once you have selected an image, you can copy the prompt and edit it in the canvas. The AI canvas provides you with a block that generates new art based on the prompt, and you can expand and change your images in ways that were not possible before.
Leonardo.ai's Inpainting and Out-Painting
Are you tired of being limited by the size of your canvas? Say goodbye to those boundaries and hello to endless creative possibilities with Leonardo.ai's out-painting via canvas feature. This game-changing feature allows you to extend your artwork beyond the initial canvas, creating a larger piece of art that truly captures your vision.
With Leonardo.ai, you have the ability to expand your images in any direction you choose. Simply select the area you want to generate new art in and press the generate button. From there, you can choose from different generations and pick the one you like the most. Whether you want a wider, taller, or even square image, the choice is yours!
And the fun doesn't stop there! You can also upload a previous generation image or create a new one and generate a full portrait by expanding and editing the image. And with the draw mask feature, you can even edit specific parts of the image – like changing the color of someone's lips – to make it truly your own.
Leonardo.ai also allows you to blend two images together. You can use the eraser to remove parts of the images that you don't want and generate a new image that blends the two together. This results in some truly amazing and unique images that you can keep adding to and changing.
Prompt Generation Tool
One of the unique features of Leonardo.ai is its prompt generation feature. The tool allows users to generate prompt ideas based on a specific word. For example, if you type in "potato" the tool will generate four prompt ideas related to,,, potatoes. This feature can be a great help for users who are struggling to come up with ideas for their artwork. It’s a simple but handy feature for those times you lack inspiration or just need some ideas to get you going. Also handy is the generate button right next to the suggested prompt that will instantly into four amazing images.
AI Pose Feature
The AI Pose feature, which seems to be an implementation of Controlnet for Stable Diffusion, in Leonardo AI allows users to generate images based on a particular pose. In order to use this feature, users need to grab a photo with the desired pose and import it into the platform. The feature requires a model trained on Stable Diffusion 1.5, such as the fine-tuned illustration version 2 model.
To generate an image, users need to provide a prompt and select the image dimensions, guidance scale, step count, and tiling options. The rendered results may not always be perfect, but the quality can be improved by upscaling the image.
In terms of training models, Leonardo AI's training for people may not be fine-tuned, but upscaling the images can lead to better results. The post-processing feature in the platform may be overbearing, especially for close-up portraits.
Performance Comparison with Midjourney
The design of Leonardo.ai and the images it produces draw comparisons to MidJourney. Both tools were successful in producing aesthetically pleasing images that accurately depicted the simple prompts provided. However, MidJourney was found to have a stronger sense of style and more visually appealing output.
When faced with more complex prompts, Leonardo.ai experienced difficulties in maintaining a consistent style, often resulting in non-cohesive images. Additionally, there was a strange glitch observed where an image generated earlier in a session was duplicated in a later prompt, despite having a different seed.
All This Is Free (As at date of publishing), but what's the catch?
One of the main restrictions of the free version of Leonardo is the limited number of tokens that you receive each day. On the free plan, you'll start with around 250 tokens, with each image costing roughly one token to generate. While this may seem limiting at first, it's important to remember that there are other features on the platform that cost more tokens.
Final Words On Leonardo.ai
Leonardo.ai is a cutting-edge artist tooling platform that offers market-leading features to give artists greater control over their generative content production. With Leonardo.ai, artists can rapidly ideate and create unique, production-ready assets from pre-trained AI models or by training their own models in just a few clicks. This artist-friendly platform goes beyond just visual assets, offering a complete generative content production studio to create a universe with infinite possibilities. Whether using an existing model or training your own AI, Leonardo.ai is fast, easy, and intuitive, making it a real game-changer for artists of all levels.
While some may consider it a "mid-journey killer," we put it to the test and compared it to the beloved mid-journey. And what did we find? This tool is truly something special.
Leonardo basically give you the core Stable Diffusion functionality in a simplistic Midjourney like interface that really allows you to focus on creating and iterating rather than getting stuck on tweaking parameters. It's definitely a workhorse app.
Sure, getting started with Leonardo.ai can be a bit confusing with their "waiting list", but once you're in, you'll be greeted with a user-friendly interface, a variety of pre-trained models, and the ability to create your own custom data sets. Not only that, but it also features an AI canvas for editing and creating art, a prompt generation feature for when you need inspiration, and the ability to blend and extend images.
In comparison to mid-journey, Leonardo.ai may struggle with more complex prompts and have a glitch or two, but it's important to remember that these two tools operate on different systems and have unique prompting styles.
So, AI Art lovers, don't be afraid to embrace change and try something new. With Leonardo.ai, the possibilities for creating amazing art are endless. Get ready to be inspired and let the creativity flow!