𝗣𝗔𝗧𝗥𝗘𝗢𝗡 𝗨𝗣𝗗𝗔𝗧𝗘 (𝗟𝗘𝗔𝗥𝗡𝗜𝗡𝗚 𝗔𝗡𝗗 𝗠𝗢𝗥𝗘)

Hello Patreons! Sorry about the short break of content, but it's for a good reason. 🥰

What's Happening?

To improve the quality and functionality of making images for you all. I have taken a lot of time these past few days to learn about making my own LoRa models to assist in making images in the future.

Why am I doing this?

When it comes to generating images, you can only do so much with an existing model that you are using. It gets to a point where the image you are trying to make doesn't generate well because the AI is left guessing on how an image should look beyond it's normal parameters. My major issue was creating images of myself where my face was presented in a profile view (side view). The model I was using would struggle generating an accurate representation of my face in that view.

Was it easy?

Simply put, no. After following many tutorials online, and making sure I did everything properly. Learning to train my own LoRa model was about to drive me crazy. I would run into many errors, had to restart from scratch many times and redownload assets over and over again. I was ready to give up on the whole thing.

I spent hours making my own datasets of images with the help of Microsoft Designer's Image Generator and thought I prepared everything properly. But, when I was about to call it quits, the magical AI Fairy came to me and granted my wish. All of the sudden, everything was working the way it should! This made me SOOOO relieved. All that work finally paid off and wasn't a waste of time.

THE RESULTS

Since this is my first time doing this, I was quite impressed with the results I was outputting. Below is a shot of how a profile view looked with the model I was using, compared to how it looks now with the LoRa I created.

As you can see, with the old model on its own, it had little knowledge on how my face would look from the side. Not very accurate. With the help of the LoRa I created, it can finally understand what a profile view would look like. This will become very handy when creating images where my character's face is looking back at the viewer during certain situations.

Now what?

For the future, this can mean a lot of things. Now that I'm finally understanding how to train my own LoRa models, there is much potential in having more control of the AI. To improve this even more, I'm also going to start learning how to work with ControlNet along with other methods. Even more, I also have the 3D model of Elsa, this will definitely help when creating image datasets in the future.

I'm actually excited that I'm learning so many new things. 😊

Poll Results

So, our last poll asked about the content being made. After 27 votes, the winning result was to create images of other favorite 3D characters out there in our media.  This means, once in awhile, we can create images of other well known 3D characters.

For example this could be:

• Mavis from Hotel Transylvania
• More Princess Anna
• Rapunzel
• Ariel
• Vanellope von Schweetz
• Female Disney Princess Characters
• And much more...

And what's even better, if there aren't much resources out there for a specific character, I can now make my own with my new knowledge of training my own LoRa models. 🤓

Final Note

I want to thank you for your patience with me as I try to learn more about AI image creating. Thanks to all my fans and supporters. It is the Holidays as of this update, my current job gets busy this time of year. So I'll do my best, with the time I have to push out more content before the month is out. I wish you all the best, and I look forward to seeing what I can create further into the future.

😉 Thanks for taking your time in reading this update.