Train Custom Models in Leonardo AI - Step-by-Step Tutorial

Table of Contents

1. Introduction
2. The Importance of Training Custom Models
3. Getting Started with Leonardo
4. Creating a New Data Set
5. Collecting Images from the Web
6. Using the Email Assistant Extension
7. Image Cropping and Resizing with bearme.net
8. Uploading and Training the Data Set
9. Generating Images with the Custom Model
   • Using Prompts and Keywords
   • Adjusting Parameters for Style and Magic Pro
10. Fine-tuning the Model with Additional Images
11. Conclusion

Introduction

Welcome to this Tutorial on how to train a custom model using Leonardo! In this tutorial, we will explore the process of training a personalized model in Leonardo using a few photographs and creating a new data set. Before we begin, let me mention that Leonardo is now open to the public, and you can access it for free with a daily allowance of 150 credits. This is more than enough to train a model and experiment with various features.

The Importance of Training Custom Models

Custom models allow you to create personalized AI models for specific tasks, such as generating unique images or mimicking a particular style. By training your own model, you have the flexibility to customize the output based on your preferences and requirements. Whether you're an artist looking to create original artwork or a content creator in need of unique visuals, training a custom model can give you the creative edge you need.

Getting Started with Leonardo

To begin training a custom model, you will need to access the training data sets feature in Leonardo. This feature allows you to upload a collection of images that will serve as the basis for training your model. Leonardo offers a free subscription that allows you to train two models per month. However, you can access and use all the models created by the community.

Creating a New Data Set

Before we can start training the model, we need to create a new data set. This data set will contain the images we want to use for training. In this tutorial, we will use images of the actress Wednesday Adams as our data set. Wednesday Adams is a famous actress, and there are plenty of images available online that we can easily Gather for our training.

Collecting Images from the Web

To Collect the needed images, we will be using a helpful browser extension called the Email Assistant. This extension allows you to extract all the images from a webpage, in our case, the search results for "actress Wednesday." You can install the Email Assistant extension in your Chrome browser by following the provided links in the video's description.

Once the extension is installed, you can simply navigate to the search results page and activate the Email Assistant. It will start loading and displaying all the available images on the page. You can scroll down to load more images, and the Email Assistant will continue to extract and display them.

Using the Email Assistant Extension

The beauty of the Email Assistant extension is that it allows you to download all the displayed images with a few simple steps. You can choose to download only the images with a resolution higher than 600x600 pixels, which is ideal for our purposes. Alternatively, you can download all the displayed images and delete the unwanted ones later.

Once you have selected the images you want to download, you can choose to save them in a specific folder. In our case, we will create a folder called "Wednesday 2" to keep the images organized. After selecting the destination folder, you can choose to download the images using a QR code. This method is convenient and doesn't require a direct download from the webpage.

Image Cropping and Resizing with bearme.net

After downloading the images, we need to crop and resize them to a standardized size of 512x512 pixels. We will use a handy tool called bearme.net for this purpose. Bearme.net is a local, browser-based tool that allows you to perform cropping and resizing tasks without the need for internet access.

To start cropping and resizing the images, simply select the downloaded images in the bearme.net tool. Then, specify the desired output size of 512x512 pixels. You can fine-tune the cropping process to remove unwanted parts of the images and ensure that the subjects are centered and visible. Once you have adjusted the settings, you can either save the cropped images individually or choose to save them as a zip file.

Uploading and Training the Data Set

With the resized and cropped images ready, we can now proceed to upload them to the training data sets feature in Leonardo. Once uploaded, we can name the data set and indicate that we want to train a new model using this data set. To ensure a successful training process, it's essential to provide a descriptive name for the model and specify Relevant categories. In our case, we will use the category "characters" to categorize the Wednesday Adams images.

After confirming the data set's creation, we can proceed to train the model. Leonardo will Prompt us to choose a model name based on the data set's name. We can also select the model's version and size, depending on our preferences and requirements. It's important to remember that the free subscription only allows two trainable models per month. So, make sure to choose the best options for your needs.

Upon starting the training process, Leonardo will Take Care of all the calculations and optimizations required to train the model. This process may take some time, usually around 40-50 minutes, depending on the complexity of the model and the number of images in the data set. It's advisable to wait for an email notification from Leonardo indicating that the model training is complete.

Generating Images with the Custom Model

Once the model is trained and ready for use, we can start generating images using the custom model. In the image generation tool, we need to select our custom model by choosing "Select custom model" and entering the model's name. In our case, we will select the "Wednesday" model.

To generate images with the custom model, we can provide prompts and keywords that define the desired image characteristics or style. For example, we may use a prompt like "portrait of a girl" and replace "girl" with our keyword "sks Wednesday" to trigger the custom model. We can also activate the prompt negator if necessary.

Additionally, we have the option to adjust various parameters, such as image Dimensions, to achieve the desired output. The image generation tool provides flexibility in terms of style and creative options. By experimenting with different parameters and settings, we can unleash the full potential of our custom model and generate unique and customized images.

Fine-tuning the Model with Additional Images

If we're not satisfied with the initial results or want to improve the model's performance, we can consider fine-tuning the model by adding more images to the data set. This process involves collecting additional relevant images and training the model using the updated data set.

By expanding the data set and retraining the model, we can refine its capabilities and generate even more accurate and tailored images. This approach ensures that the custom model becomes a powerful tool that aligns with our creative vision and meets our specific requirements.

Conclusion

In conclusion, training a custom model in Leonardo opens up a world of possibilities for creating personalized and unique AI-generated images. By collecting the right set of images, training the model, and fine-tuning it as needed, we can unleash our creativity and produce stunning visuals. This tutorial has provided a step-by-step guide on how to train a custom model in Leonardo, from gathering images to generating the final images using the custom model. So, go ahead and explore the potential of custom AI models in your creative endeavors!

Highlights

• Train a custom model in Leonardo to create personalized AI-generated images.
• Use the Email Assistant extension to extract and download images from the web.
• Crop and resize images using bearme.net for a standardized training data set.
• Upload the data set to Leonardo and train the model using the training data sets feature.
• Generate unique images with the custom model by providing prompts and keywords.
• Fine-tune the model by adding more images to the data set for improved performance.

FAQ

Q: Are there any limitations on the number of custom models I can train in Leonardo? A: Yes, the free subscription allows you to train up to two models per month. However, you can access and use all the models created by the community.

Q: Can I adjust the parameters of the custom model to achieve different styles? A: Yes, you have the flexibility to adjust various parameters, such as image dimensions, to achieve the desired output. This customization allows you to explore different styles and creative possibilities.

Q: Can I fine-tune the custom model with additional images? A: Absolutely! Adding more images to the data set and retraining the model allows you to improve its performance and generate more accurate and tailored images.

Q: How long does it take to train a custom model in Leonardo? A: The training process can take around 40-50 minutes, depending on the complexity of the model and the number of images in the data set. It's essential to be patient and wait for the email notification indicating that the training is complete.