Combining LoRAs is one of the most effective techniques for achieving consistency in characters and styles. Plus, it's super fun! Want to know how? I’ll explain in the comments 🧵👇 (Bookmark this for later)

First, you’ll need your trained LoRAs. I’m combining my custom LoRA (with my face) with my LoveWordly and MandaPunk LoRAs to generate versions of myself in these styles. If you haven’t made yours yet, here’s the step-by-step guide:

In the previous tutorial, you saw how to train LoRAs and how to extract the link for the .safetensors file. We’re going to need that, because from @FAL, we can add one LoRA to combine with the one already selected. Remember, you need to use the trigger words—both of them!

The prompt: A bald baby with a beard, techhalla_flux, in a diaper holding a laser gun | alone in a vast desert surrounded by cacti | playful yet surreal scene | bright, dry landscape with scattered plants | in the style of mandapunk

Techhalla_flux is the trigger word to generate my avatars, and mandapunk gives them that unique style. From there, create your prompts and have fun! Here are some examples of mandapunk and loveworldly LoRAs.

You’ll notice there’s a "Scale" parameter under each LoRA. It allows you to apply more or less weight to your LoRA compared to the base model, in this case, Flux 1.1. I encourage you to play around with it, but from my experience, try to keep the total of both scales below 2.

If you want to try my LoRAs, don’t forget you can do it on @getremixai! Here’s all the info:

And finally, you can animate your results with @Kling_ai, they're absolutely insane! I’ve got a GPT to generate optimized prompts for Kling. If you need it, just ask!

I’d love to hear what you think. And if you enjoyed it, follow me at @techhalla to stay up to date with all the generative AI updates, and don’t forget to leave a like, repost, and comment on the main post. It really helps me keep sharing these contents!

The master of Ai Tools!

thank you buddy!