Trained on ~80 album covers, mostly from the 50s and 60s, a mix of Jazz, pop, polka, religious, children's and other genres.
Sample Prompts:
Kanye plays jazz, albumcover style
Swingin' with Henry Kissinger, albumcover style
Jay Z Children's album, albumcover style
Polka Party with Machine Gun Kelly, albumcover style
Star Wars, albumcover style
Description
FAQ
Comments (7)
please share the training data
I added the instance images to the repo on huggingface -- https://huggingface.co/lckidwell/album-cover-style/tree/main
This looks good! How did to train this? Did you use DreamBooth to fine-tune SD?
Is this UNet? (Sorry for silly questions, I'm new to SD, came from GANs). I see the sizes of models shared by different people vary. For instance, https://civitai.com/models/5001 is just 144 MB but yours is in GBs? How?
@rahulbhalley This is a dreambooth model - I tried to make an embedding of it (just a few kilobytes) but didn't turn out well. The link you shared is to a safe tensors file, another way to apply weights to an existing model
if you're using stable-diffusion-webui, put the 144mb ones in your /models/Lora folder
@muddylemon To me stable-diffusion-webui seem very abstract so I skipped trying it for now. But I'll surely try it after I learn to fine tune SD with python script.
Btw, how long does it take to train a LoRa model and is its quality close to original fine-tune scheme?
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.