Join me on Patreon for exclusive Pro Models, workflows and Lora's and early access to unique resources. To commission LoRa's or checkpoint, feel free to connect on Discord.
Onsite generations are available on these models:
👉 Asian_Realism_By_Stable_Yogi (PONY) : https://civarchive.com/models/168949?modelVersionId=1624625
If there’s another model/version you’d like to generate images with, just place a bid here and I’ll make it available for onsite generation:
https://civarchive.com/auctions/featured-checkpoints
👍 Like this model to keep me motivated and inspired to create more!
💬 Drop a comment and let me know what you'd love to see next.
🌟 Review this model to help me improve and make even better creations.
🔔 Hit that notification bell to stay updated with my latest models and updates!
🌟Buzz Buzz Buzz, we need buzz to continue paying electricity bills.
Important Usage Tips
Add Stable_Yogis_PDXL_Positives at the beginning of your prompt section.
Add Stable_Yogis_PDXL_Negatives-neg at the beginning of your negative prompt section.
Why?
The difference is enormous—check the image below for same-seed comparison with and without these Embeddings. Link to comparison images https://civarchive.com/images/41648855
For Best Results use My Negative Embedding with Detail sliders and Checkpoints
Description
Asian_Realism_pony_V1_VAE_FP16_BY_STABLE_YOGI
Pro Version of Most of my models are now available on My Patreon
This Checkpoint is Better with Asian faces and anatomy also works great in realism, Skin Details, Lighting, dynamic poses, diverse faces and its . Please test and Share feedback.
Sampler Euler a,
CFG Minimum 4+ maximum upto 7
Steps Minimum 18 , Recommended Steps 27
Resolutions: All SDXL Resolutions
Adetailer (Required)
High res Fix (Optional)
Denoise - 0.30
Minimum hires steps 5+
Minimum Upscale 1.5
Upscaler 4x-UltraSharp
Keep the VAE at Auto ( VAE Included )
Negative Prompts
(Download this negative embedding for best results) Stable_Yogis_PDXL_Negatives
Positive Prompts
(Download this positive embedding for best results) Stable_Yogis_PDXL_Positives
FAQ
Comments (11)
Any plans to add a DMD2 version of v1.0 or a version based on Illustrious?
DMD2 version Ready and landing on 11th
Illustrious in the works, will release soon.
@Stable_Yogi Thanks for the update. I'm looking forward to it.
Is there a significant difference between FP16 and BF16? I have run all the models in BF16 but I never stopped to think if quality or accuracy is lost
As per Chatgpt Bf16 has more accuracy. I am still testing this , not really sure if that is the case. But your test results can help.
Thanks
@Stable_Yogi From what I could understand, BF16 works better when you downgrade from FP32, for example, quantizing an FP32 model to BF16 will have more numerical precision than from FP32 to FP16. However, this benefit only exists in this case, when you convert from a source in FP32. I don't understand very well the training part of the image models, but if you are using sources that are in FP32, BF16 is really the best option. I don't know if these base models are originally made in FP32, I know that SDXL was designed to run in FP32, but I don't know if the Pony and Illustrious base models were also made in FP32 or at least in BF16. I think there have been so many merges that it no longer makes any difference. The ideal would be for someone to do training entirely in FP32, or BF16 and forget about FP16. This way we would have much more accurate models, but I think that at this point, redoing an entire base model like Pony or Illustrious (if they weren't made in FP32 or BF16) wouldn't be worth the expense, or at least there wouldn't be anyone willing to do it
，
If there’s another model/version you’d like to generate images with, just place a bid here and I’ll make it available for onsite generation:
https://civitai.com/auctions/featured-checkpoints
Can you also train custom Lora just for me? If yes how much would you charge?
@tempmailzorg591 sure i can do that.
join me on discord to discuss. https://discord.gg/FAgfAeMqys
Cool