→ Get it now for much cheaper prices, skip the wait:
🛒 Ko-fi · TensorArt ✅ Card & PayPal accepted — no crypto, no Buzz, no queues ✅ Instant download · Priority support
🛒 More models in my store → 🛒 Whop · Ko-fi ✅ Card & PayPal accepted — no crypto, no Buzz, no queues ✅ Instant download · Priority support
💎 Get the complete Event Horizon collection → VIP on Whop · Ko-fi
I wanted to make an specific embedding for John William Waterhouse and Alma Tadema among others painters (i was honestly surprised there's nothing in civitai about them or couldn't find anything related to) ... so I've decided to mix everything because "Western Diffusion" is already taking me so much time.
It makes amazing landscapes also.
The dataset comprehends a thousand of pictures with "similar aesthetics"
Feel free to suggest more classic artists styles (but please no Alphonse Mucha, Gustav Klimt or Greg Rutkowsky)
If you like my embeddings, please consider to support my work in
Description
FAQ
Comments (11)
What base model did you use to train this? I've been wanting to train Classical styles but I'm not sure which model to use to train them.
Hi! I just use SD 1.5 pruned emaonly safetensors for everything atm. I think is pretty much the standard procedure (it's compatible with almost everything). I'm a rookie anyways, so don't take my word for it.
@Punkimaster Awesome :) I've trained other styles using 1.5 and they came out pretty well, so for styles that aren't anime I'll stick to that. Great advice!
@madaraxuchiha88 You're welcome.
@Punkimaster When creating new models they should be trained on the original Stable Diffusion 1.5 pruned.ckpt or the 1.5 pruned.safetensors.
That way they can be used with ALL checkpoints without problems.
It is the 7.7 Gb file you can find here - https://huggingface.co/runwayml/stable-diffusion-v1-5/tree/main
When training art Styles, the models should have at least 50 to 75 images (or more) for training and should be tagged with the words "in the art style of ____name of artist___" in the BLIP-caption files.
Train ONE Artist style in the same model if you want proper AI created artworks, otherwise the AI will start mixing other art styles into your images.
If you train your Models using a Civitai Checkpoint the models will only work best if you use that particular Checkpoint, which may not be the one everyone else is using, and the images will come out BAD.
@maxhitman thanks! I just use that model because I read it consumes less VRAM. My specs are less than the recommended for training so I try to keep the VRAM use as low as I can. I've tried my embeddings with the most downloaded models here and I didn't have any trouble regarding the fidelity of the output so far.
This is pretty fantastic! Have you considered trying your hand at a Patrick Negal model? I think you would knock it out of the park.
There's already one in Civitai, an it's pretty good so it would be a little bit redundant. https://civitai.com/models/55013/natrick-pagel
I struggle with training textual inversion, are you following a particular guide? any resources you could point me to for TI training? or maybe you could share a tutorial someday, thanks
Hi! I'm not an expert. I've just start training two weeks ago. I just followed a tutorial from civitai about embeddings.
@Punkimaster can you link me to it?
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.