Isometric Dreams
Trained on ISOMETRIC styled images from midjourney that we got from a friend of ours.
This is now also up on TensorArt: https://tensor.art/models/635481845263489518
MONTHLY & ONE TIME SUPPORTERS GET PRE RELEASE ACCESS & MORE:
WE ARE PROUDLY SPONSORED BY: https://www.piratediffusion.com/
If you got requests, or concerns, We're still looking for beta testers: JOIN THE DISCORD AND DEMAND THINGS OF US:
JOIN OUR SUBREDDIT: https://www.reddit.com/r/earthndusk/
Listen to the music that we've made that goes with our art:
https://open.spotify.com/playlist/00R8x00YktB4u541imdSSf?si=b60d209385a74b38
Description
If you want to donate towards costs and don't want to subscribe:
https://ko-fi.com/DUSKFALLcrew
If you want to monthly support the EARTH & DUSK media projects and not just AI:
FAQ
Comments (14)
Which YAML is needed to run this?
None ❤️❤️❤️
I didn’t know there were 2.x models that didn’t need a config. Does that mean they just use the default config?
@justmaier I DONT KNOW? XD The only models i've ever used requiring config was project unreal and any merges i made after i used that model.
Doesn't work with latest A1111 at all. Just what are you running it under?
RuntimeError: Error(s) in loading state_dict for LatentDepth2ImageDiffusion:
size mismatch for model.diffusion_model.input_blocks.0.0.weight: copying a param with shape torch.Size([320, 4, 3, 3]) from checkpoint, the shape in current model is torch.Size([320, 5, 3, 3]).
@Baughn I trained it on huggingface, used it on A1111, but maybe the verison i ran did work ? Not sure, did you try the 1.5 model?
@duskfallcrew What 1.5 model? It still says 'coming soon'?
That being said, TEST The 1.5 and then please don't give it a 1 star i'm aware it's rabies infested with watermarks, i made a stupid mistake XD
1.5 Version being trained ASAP as someone wants me to merge it with theirs :P
Why would a model even need a trigger word (and it does)... it makes no sense.
Why would you comment on something that makes no sense, plenty of models have this - it was trained at huggingface. Using dreambooth a lot of older models need this - it's not a merged model it's a trained one. Maybe newer models don't need it, but this was months ago so SHRUG* - I can't really fix this issue so?
This is overkill for the isometric style.
A Textual Inversion can achieve the same outcome, whilst retaining the models generalization ability and can be trained in significantly less time.
I suggest you use your training images to train a Textual Inversion instead and see how that comes out.
The preview images are also lacking in fine detail, this can be due to the training images being a bit blurry themselves, make sure they are either very excellent images, or upscale them.
grins You do realize this model was trained WAAAAY Back in like Feb? When I DID NOT know what I was doing and had VERY EARLY on midjourney data?
<3 I've had better ones made since then, and I fully appreciate your comments btw, I'm not knocking you at all - I don't usually train TI's - i did train Iso TI's but frankly, it's fun to do whatever works <3
This was one of my EARLIEST and probably only one of a few 2-2.1 models, so if you look up more recent ones like Paradox Isometric or Insane Isometric :)