How to train AI with your face

In partnership with

If you are on social media, you may have noticed that some people is sharing AI-generated images of themselves on, like crazy things with his face

Like this one

You’ve probably wondered: How can I do that?

The answer lies in something called LoRA’s.

Let’s break it down: what are LoRA’s, and how can they help your brand stand out?

Learn AI in 5 minutes a day

This is the easiest way for a busy person wanting to learn AI in as little time as possible:

1. Sign up for The Rundown AI newsletter

2. They send you 5-minute email updates on the latest AI news and how to use it

3. You learn how to become 2x more productive by leveraging AI

Sign up to start learning.

What Are LoRA’s?

LoRA’s (Low-Rank Adaptation) are a smart way to personalize AI models, like Stable Diffusion or Flux, without starting from scratch.

Training an AI model from zero can cost thousands of dollars, weeks of work, and require powerful computers.

For small businesses, this isn’t practical.

LoRA’s solve this by letting you train an existing model with just 3-10 images of you, your product, logo, or anything else you want. The model then uses this to create consistent and personalized images.

Why don’t we train models ourselves?

• Most of us don’t have access to the high-powered GPUs needed.

• It’s technical and time-consuming.

How Can LoRA’s Help Your Business?

1. Social Media Content

Imagine being able to create stunning images of your products in styles that suit any occasion.

Picture your brand image or your products on a sandy beach, surrounded by festive lights, or styled for a seasonal sale like Black Friday or Christmas.

You get tailor-made images for every campaign with minimal effort. 

2. Product Prototyping

Have new product ideas? Instead of investing in expensive prototypes, you can use LoRA’s to visualize your designs first.

Test out different colors, packaging, or even entirely new concepts without committing to production. It’s a faster, cheaper way to explore your creative options.

3. Brand Consistency

Keeping a consistent look across all your platforms is key to building trust and recognition.

LoRA’s make this easy by generating images that feature your logo or specific product visuals every time. No matter the context, your brand identity shines through in everything you share.

How to Train a LoRA

When it comes to training a LoRA to generate personalized images, there are several methods to choose from, depending on your resources, technical expertise, and goals

Let’s explore the main options, along with their advantages and disadvantages

1. Online Platforms

• What It Is: Services like Freepik allow you to upload images and train a LoRA directly on their platform.

  (It’s the one that I used to create the first iamge)

• How It Works: You provide your images and the platform handles the heavy lifting. Only for faces so far.

• Advantages:

  • No need for technical expertise or powerful hardware.

  • User-friendly interfaces with guided workflows.

  • Typically fast, since platforms optimize their infrastructure for this purpose.

• Disadvantages:

  • Cost: Many platforms charge per project or require subscriptions.

  • Privacy: Your data is processed on their servers, which may raise concerns for sensitive projects.

  • Customization: Limited control over the model's architecture or parameters.

2. Local Training on Your Own Computer

• What It Is: You download tools like Automatic1111 or ComfyUI to train the model locally on your own hardware.

• How It Works: Using your GPU, you configure the training process by loading your dataset and specifying parameters.

• Advantages:

  • Complete control over the training process and settings.

  • No recurring costs once hardware is acquired.

  • Ideal for those prioritizing data privacy, as everything is processed locally.

• Disadvantages:

  • Hardware Requirements: Training a LoRA requires a powerful GPU with sufficient VRAM (at least 8GB, though more is better).

  • Complexity: Requires technical knowledge to set up, install dependencies, and troubleshoot errors.

  • Time: Slower on consumer-grade hardware compared to cloud services.

3. Cloud-Based Training with Rented Computational Power

• What It Is: Services like Google Colab let you "rent" GPUs in the cloud to train your LoRA without needing local hardware.

• How It Works: You upload your dataset, set up the training environment, and run the training process on a virtual machine.

• Advantages:

  • Flexibility: Access high-performance GPUs for a fraction of the cost of buying one.

  • Scalability: Increase computational power as needed.

  • Ease of Setup: Pre-configured environments and tutorials simplify the process.

• Disadvantages:

  • Cost: Renting GPUs can add up quickly, especially for long or repeated training sessions.

  • Learning Curve: Some platforms require familiarity with coding or command-line tools.

  • Dependency on Internet: Training requires a stable internet connection.

As you can see, this week there is no AI Filmmaking Challenge

The project just started and I am trying new things, but the most important thing didn’t change

I want you to learn how to create images and videos without experience

So let me know, what are you missing?

What would you really like to learn?

Just reply to the email

Bigger changes are coming, don’t miss the opportunity to make an impact on the project