ƤŘØΜƤŦƤƗŘΔŦ€
Ahoy 🦜🏴☠️
[⚠️WARNING !!! MAY GENERATE PHALLUS/VAGOO SHAPED OBJECTS!!!⚠️]
Took a break from voice models to test out a new GPU with Z-Image lora training.
This one is 5000 steps trained on Necronomicon, Alien Artwork, and assorted H.R Giger collections.
Came out much better than I'd thought imo for a first run. It makes very nice soft tones, and works well at high settings. On lower settings it gives things a hint of that lube-like H.R.Giger sheen
No trigger word needed👌
Example: (WITHOUT GIGER LORA)
Example: (WITH GIGER LORA)
Edit: after a ton of testing, the one thing I'd like to fix asap is that it generates images that are just a little bit too bright. Going back through the dataset I've realised that many of the images aren't as dark and slimy as I thought, so presumably that's it.
It's nothing post processing can't fix anyway, but I'll improve for another attempt.
I follow and sub to this anon SECourses Creator Profile | Civitai if you're interested in setting up Z-Image (or are setting up almost any AI repos in a doofus proof way)
Z-Image Turbo LoRA training with AI Toolkit and Z-Image ControlNet Full Tutorial for Highest Quality | Civitai
Another good anon to follow is Aitrepreneur | Ai art, Large Language Models, Stable Diffusion, LLAMA, LORA | Patreon
Not me I promise, just paying it forward for how much help these two have been