DeamPony Model Overview
A Frankenstein-style merge of various pony models, including recent PonyRealism additions. This was a trial and error merge process with no technical knowledge whats so ever,
Recommended Parameters for Generation
These are the samplers I have used and have had success with, it does not mean others wont work please experiment.
Samplers:
*Recommended:**
DPM++ 2M Karras*
DPM++ 3M SDE Exponential*
DPM++ SDE Karras*
Settings:
*Steps:** 25-35
*CFG:** 2-5 (for realism) Higher the 5 things can get a little odd but feel free to experiment
*Clip Skip:** 2
*Resolution:** 1024x1024 | 1024x768 (or similar variations)
Additional Information:
Use Danbooru tags for best results.
Standard SDXL prompting also works well.
These are my usual default prompts
Recommended Positive Prompt:
(score_9:1.2), score_8_up, score_7_up, absurd_res, hi_res, detailed face, highly detailed, very aesthetic, Cinematic, 4k, 8k, absurdres, RAW, source_photo, high quality photography, Canon EOS R3, golden hour, 50mm, f2.8, natural light, {prompt}
Recommended Negative Prompt:
score_6, score_5, score_4, censored, close-up, monochrome, mutated hands, low res, blurry face, (deformed, distorted, disfigured:1.3), poorly drawn, bad anatomy, ugly, disgusting, blurry, watermark, artist signature, artist name, worst quality, low quality, human, anthro, drawing, painting, crayon, sketch, graphite, impressionist, noisy, source_anime
Note: Due to the experimental nature of this model, results may vary significantly between generations. Embrace the unexpected! Further experimentation with this latest version may yield optimal results for different parameters.
Description
FAQ
Comments (16)
Excuse me in advance for the noob question,
But how am i supposed to use the model?
Do i have to download it?
Yes you have yo download it and use it with a program like automatic1111, or comfyUI
@SD1one oh ok,
Well it is becoming a passion to generate ai stuff but im still mobile based, maybe in a while haha.
@Falkirves I actually run a program on my computer which I then access from my phone to do most of my work.
@SD1one how do we get civitai to have this available in its onsite generator? its incredible - anything i can do like contact them?
@Legend_of_Chavo_G honestly I have no idea
@SD1one I would suggest you ping support to enable the model.
> Thank you for reaching out to Support! I apologize for the inconvenience. When creators upload their models, they have a section dedicated to permissions for that model, where they can select 'Use on Civitai generation service'. Usually, this will work to populate the blue Create and Download buttons for the model page and enable on-site generation, but sometimes, an error occurs and it remains solely as a downloadable resource.
When this happens, the creators can reach out to Support and provide the links for each of the affected models. We will be more than happy to manually add this functionality for their model(s).
I hope this helps!
Best regards,
Jane
@ffc4now thanks
@ffc4now civitai has informed me they have no more space for new models at this time
@Legend_of_Chavo_G civitai has informed me they have no more space for new models at this time
tried in fooocus and comfy with many different settings flows configs and all of them generate the image fine in preview but when saving it seems to try add some post processing that makes the colour washed out and covers in in white dot artifacts, image below is great though - only happens with this model
Hmm does it having something to do with the VAE?
Sounds like a VAE issue, I had a similair issue with ponykaine
@SD1one @A_Friendly_Spider Ah yeah you guys are right, swapping out the VAE did it! thanks!