Here is the Lora for body writing !
The trigger word is body_writing
Sometimes, there may be speech balloons in the pictures. You may add "speech balloons" in the negative prompt to reduce the chance.
Hope you enjoy my work!
You are welcome to browser my other works if you are interested!
#
New Announcement (19th April)
Hello my followers. I am sorry that I bring a bad news to you today. The update frequency of Lora may be dropped dramatically or even be stopped for some reason. You can still leave your request for free (discord: qqqw123#4777) but I am not 100% sure I will make it. Thanks for your support and appreciation!
#
You are also welcome to leave your request for free and I will try to make it into a Lora if it is possible.
If you appreciate my works, you can support me via donation on ko-fi.com/dpp12
Description
FAQ
Comments (8)
This is somehow ironic, considering that AI cannot write text. :-D I guess you will eventually need to paint in the writings and use ControlNet or Photoshop to write real ones. :-)
Actually, the result of the Lora cannot meet my original expectation. I want to demonstrate the japanese text and I used those pictures to train it. Anyways, I think this Lora can only be a substitution. Thanks for your suggestion!
@dpp12 it won't do any text, even if you train it on real hieroglyphs it won't reproduce them, just scribbles that look like them as any other language. AI can't draw text properly cuz it doesn't regards them as words, just as a pattern so it makes the overall feel of them not the exact words. :-)
@drakmour OK. I understand what you mean. Thanks for your information!
yeah i wish this was trained with english but i was thinking the same thing if you used images with common phrases and what not or specific phrases you want you could train the ai to write them or more specifically draw the writing I suppose but not all ai is incapable of generating text so your comment is kinda wrong just not abt this one they just have to be built to process information in that way which isnt the case here and would need a rework of stable diffusion itself.
that said i handwrite mine in currently by mouse thinking of getting a drawing tablet
This is the only lora I have issues with. Whenever I use txt2img or inpaint with this, it gives me this error:
AssertionError(f"Bad Lora layer name: {key_diffusers} - must end in lora_up.weight, lora_down.weight or alpha")
AssertionError: Bad Lora layer name: lora_unet_down_blocks_0_downsamplers_0_conv.lora_mid.weight - must end in lora_up.weight, lora_down.weight or alpha
Any idea?
Because this is not a LoRA bro.