Hi again! Long time no see.
Welcome to Coinmaker. Coinmaker is my second "concept art generator" here. This time I trained a coin-asset-generator LORA for SDXL 0.9 that will makes easier, more fun and aesthetics the coin assets creation for videogames, renders, etc. It's specially useful for 3D projecting and extrude-modelling in 3D softwares. So that's why instead of having an small coin it will generate a big coin to be able to extrude all details (as requested)... also he told me that could be used for logo/shield but that's not the main purpose.
Token is c01n.
IMPORTANT: This was tested under ComfyUI, I don't have any idea how it will go with Automatic1111. I strongly recommend you to use ComfyUI if you want good results and most important, if you want the same settings I used just drag any image into the ComfyUI and it will show the full workflow, it seems CivitAI converted all samples to JPG so instead of dragging the images (if the workflow is not appearing):
https://huggingface.co/datasets/davizca87/C01nWorkflow/tree/main <- LOAD it via Json file in Comfyui
How to prompt and which values to use?
********************************************
- The LORA works perfectly with values from 1-1.2 and bellow even IMO it's better to stay at 1 (you can see Clip and Model 1/1 values).
- Resolutions 768x768 and 1024x1024 onwards, very important the ratio is a SQUARE.
- You should prompt with a little description of what you want having in mind it was trained to have a clean background and a type of metal. This is just for an specific purpose but currently support 4 metals (gold, silver, bronze, ancient).
- Structure as: "photograph/design/artwork of a (type of metal) c01n with a concept-you-want inscription over black background". Black background makes cleaner creations while inscription even optional push forward the trained aesthetics.
This LORA is good for
**********************
- 3D renders of coins of different type and styles. <- Main purporse
- Variations of a lot of coins inner properties. ex: bloody, melt, drip, damaged...
- Concepts engraved in the coins. Some may need prompt tweaking
- Realistic or semi-realistic results.
This LORA is not to be used for:
**********************************
- Moving the coins between or across the scenes or putting them inside any recipient.
- Making ANY figure/character engraved, some won't work. Thats' AI limitations.
- Fill an scenery of coins.
Massive thanks to the original user who requested this LORA and let me upload this to CivitAI so all could enjoy. Hope you all enjoy this and thanks for the continue support in our Patreon
Have fun! David.
Description
v1.1 - SDXL 0.9 version ported
FAQ
Comments (14)
Sounds like an interesting concept...will try soon!
Thanks, feel free to throw your feedback. If there is room to improve I will dig into it and since SDXL is new I'm sure there is a lot of territory to explore.
How do I load the loras? Anyone have a workflow? Dragging an image from civitai does not load the workflow. TIA
It seems CivitAI converted the files to JPG, gonna fix it.
Edit: Post fixed. Link for workflow:
https://huggingface.co/datasets/davizca87/C01nWorkflow/tree/main <- Download the Json file and load it.
Edit2: in the case you want to see all images workflow (which should be the same), I have added a public link in our Patreon to a .rar file with all the original files untouched (PNG's).
I am using Automatic1111, the 25.07.23 update supports SDXL as well as Lycoris and Loras for it. No ControlNet yet though.
@Daedalus_7 Keep in mind that Auto1111 and Comfy will spit different results cause the way the seed is generated. But other than that, shouldn't be too much of a problem except if you want to replicate generations.
Guess I have a reason to consider downloading Comfy UI. Do you need to go through a long process to install it locally? Also, the checkpoint model used, could you drop a link to that as well? These automation of design would be an interesting thing to experiment with. Not only as it would be a great material for the book I am writing, but the coins already generated look so damn cool.
Sure. I used the base 0.9 and refiner, both official from Stability's huggingface.
About installing is very easy, I installed a portable ComfyUI version from the github I posted above.
@LDWorksDavid I went to the huggingface and it required me to submit something about what research team I was apart of. The thing is that I am just a private person that like the results generated, that's all.
@MiniMage SDXL requires authorization from Hugging, same as SD 1.5, 1.4, etc. Just fill the things with whatever you want. It's not important. Put any company name, etc.
Tutorial: https://www.youtube.com/watch?v=FnMHbhvWUhE&t=951s
Works well with the latest version of A1111. I just dropped it into the Lora model folder and used it with the SDXL Dreamshaper alpha. Works fine. Thanks for making this.
You welcome!
pls 1.5
Will consider it but right now I have some patreon request in queue and those will come first.
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.