In sukatto videos, there are often scenes where a character like a 'bad mom friend' steals something on her own, uses it incorrectly, and then yells at you through the intercom, saying, 'It's your fault I had a terrible experience!' I've created a LoRa for that scene. Mainly at 0.4, an image of the target character is generated on the intercom camera, but it may not work well, such as coming out of the frame or being in the foreground. Also, if you strengthen it to 0.6, an image that fits within the screen will come out, but the drawing is fixed, and there are parts where the style of the target character is broken. So please keep generating images until a good one comes out where the target character is correctly displayed.
Description
FAQ
Comments (7)
Great job, do you do your trainings locally or you use Google Colab? I'm building my first dataset myself.
in my case, Google Colab. i don't have enough money for local
Interesting, I'll have to use it too, my graphics card isn't good enough at the moment for local
@goldhopper is there any way I can chat with you more effectively? I'm creating my dataset for a Lora training, but what I want to do is very specific, is not a character, or style etc. I have just some few questions like if it would better to train a lora or to create a dreambooth model altogether, and about what kind of images I should have or not in the dataset to achieve my goal.
@izimontana1683
Thank you for reaching out with your questions about creating a dataset for LoRa training. I'm currently involved in ongoing research with LoRa myself, and I constantly adjust my datasets and settings based on new developments in LoRa technology. However, I'm currently facing a specific challenge: I'm trying to generate a LoRa with a keyhole embedded in the body, but unfortunately, the keyhole is not being recognized at all in the Stable Diffusion model, leading to some frustration.
So, perhaps, you had better ask with reddit. I hope it's helpful.
@goldhopper good luck with your LoRa keyhole .
cool lora !! thanks for it!
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.