- The Signal
- Posts
- How to Train an AI on Yourself
How to Train an AI on Yourself
E-commerce and product photography will never be the same...
Roberto Nickson
August 28, 2024
Welcome back to The Signal, a weekly letter where I share stories, trends, strategies and insights to help you level up as a creator and win on the internet.
Today’s topics:
How to Train an AI on Yourself
Powerful Websites Every Creator Should Know
10 Things to do Today to Grow on Instagram
As always, feel free to respond directly to this email with your feedback, suggestions, or just to say hi!
How to Train an AI of Yourself
You can generate an image (and animate it if you’d like) of anything you can imagine with yourself as the main character.
To me, this is a huge deal. It’s something that I think will have a dramatic effect on everything from YouTube thumbnails to e-commerce.
The reason being is that you can also train AI models on specific objects.
For example this is my friend Linus generating a YouTube thumbnail of himself wearing the Ray-Ban Meta smart-glasses.
Now there are a lot of ways to do this.
The best and most accurate way at the moment is to fine-tune a FLUX model. This blog post is the most straightforward explanation I have come across.
However, it takes a bit of technical ability, so I will focus on a much easier way to accomplish this.
We’ll be using Letz.ai.
I have no paid affiliation with this product. There are several like it, but I found Letz one of the easiest to use with the best results.
I actually made a video on it which ended up crashing their servers, but it looks like they are back up and running now :)
Step 1:
Head over to Letz.ai, and create a free account.
Once you’re in, click on “Create AI Model” at the bottom right.
Step 2:
You can make your model public (allowing others to use it) or private, for only you to use.
Now, enter a name for your model. This will be needed to call the model for prompting.
Step 3:
Choose whether your model is a person, object or style – meaning a visual aesthetic.
Upload 5-10 images of yourself. Follow their guide for best practices here. The better quality and detailed the images, the better results you will get.
Step 4:
That’s it! Wait for your model to be done training, activate it, and start prompting.
Experimenting with your prompts and settings to get the exact look that your’e after.
You have to @ mention the model you’ve trained to include it in the image generation.
Step 5 is optional. But you can then run your image through Runway or Luma’s image-to-video model to animate it.
You can see that in action here.
10 Things you can do Today to Grow on Instagram
I filmed a new YouTube video with 10 of my best tips for growing on Instagram.
I’m going to be making a lot more long-form videos so I would appreciate you subscribing!
Powerful Websites Every Creator Should Know
I made an Instagram carousel featuring some websites and tools that are all part of my creative process.
Here’s how it was designed in Photoshop (and completed in Premiere)
Content Resource of the Week
Check this folder out.
It includes some great templates for the type of viral reels I described in a past issue of this newsletter.
Quick Hits
This section includes things I have found interesting and helpful this week.
‣ Matt explains how to train a LORA using FLUX.1 with your own likeness. I tried the workflow out myself and it’s pretty easy with great results. Another alternative to the two mentioned above.
‣ Linus explains how AI training featured above is going to change e-commerce forever.
‣ Kevin shows off a very clever way to create VFX with generative AI.
