uwumerge (Furry Checkpoint)

( Not my Model )


 <a href="https://civitai.com/models/2368916?modelVersionId=2767854" rel="noopener noreferrer nofollow" target="_blank">
 https://civitai.com/models/2368916?modelVersionId=2767854
 </a>




 About this version


 v3 is built on a blend of Illustrious + Chenkin-Noob models, I transferred and strengthened the style by extracting uwumerge v2 as a LoRa and patching it onto the models alongside uwustyle v4. Also now has Neptunia XL VAE baked-in.


 The result should be a bit better in terms of colors, backgrounds, details and prompting.

v3.0

( Not my Model )


 <a href="https://civitai.com/models/2368916?modelVersionId=2711935" rel="noopener noreferrer nofollow" target="_blank">
 https://civitai.com/models/2368916?modelVersionId=2711935
 </a>




 About this version


 This version is
 <code>
 uwumerge v2 + DMD2 * 0.65 + SDXL-Flash * 0.55
 </code>


 Optimized for 6 to 12 steps and CFG of 1.0 to 2.0, but quality will be lower than v2.0 in some cases




 Checkpoint version of uwumerge, based on a blend of Illustrious &amp; NoobAI.


 The default style emphasizes strong outlines and colors.


 
 Fast
 
 versions are optimized for low CFG and 6-12 step generation. I use a blend of DMD2 and SDXL-Flash that I think preserves the style well, but there will still be differences from base versions.


 
 FP8
 
 pruned model stores the UNet and CLIP models in FP8, but the VAE in FP16. This reduces the byte size by almost half in return for a slight quality loss, but most likely won't support hardware-accelerated FP8
 
 inference
 
 (so it won't generate faster).


 
 Tips:
 


 Start with Euler A and a CFG between 5.0 - 9.0 (if using v1.0 then use 4.0 - 7.0). DPM++ samplers can be good too, but CFG range may vary.


 For step count I typically use the formula
 <code>
 5 + cfg * 4
 </code>
 , but 25 is also a good middle ground.

v2 Fast (CFG 1-2)

( Not my Model )


 <a href="https://civitai.com/models/2368916?modelVersionId=2687142" rel="noopener noreferrer nofollow" target="_blank">
 https://civitai.com/models/2368916?modelVersionId=2687142
 </a>


 
 About this version


 v2.0 is built on the new improved uwustyle v4.0.


 I think this version has more personality and depth.


 This version also handles CFG above &gt;7.0 better :)




 Checkpoint version of uwumerge, based on a blend of Illustrious &amp; NoobAI.


 The default style emphasizes strong outlines and colors.


 
 Fast
 
 versions are optimized for low CFG and 6-12 step generation. I use a blend of DMD2 and SDXL-Flash that I think preserves the style well, but there will still be differences from base versions.


 
 FP8
 
 pruned model stores the UNet and CLIP models in FP8, but the VAE in FP16. This reduces the byte size by almost half in return for a slight quality loss, but most likely won't support hardware-accelerated FP8
 
 inference
 
 (so it won't generate faster).


 
 Tips:
 


 Start with Euler A and a CFG between 5.0 - 9.0 (if using v1.0 then use 4.0 - 7.0). DPM++ samplers can be good too, but CFG range may vary.


 For step count I typically use the formula
 <code>
 5 + cfg * 4
 </code>
 , but 25 is also a good middle ground.

v2.0

( Not my Model )


 <a href="https://civitai.com/models/2368916?modelVersionId=2664111" rel="noopener noreferrer nofollow" target="_blank">
 https://civitai.com/models/2368916?modelVersionId=2664111
 </a>




 About this version


 Uses vaemix v6 as the built-in VAE model.




 Checkpoint version of uwumerge, based on a blend of Illustrious &amp; NoobAI.


 The default style emphasizes strong outlines and colors.


 
 Fast
 
 versions are optimized for low CFG and 6-12 step generation. I use a blend of DMD2 and SDXL-Flash that I think preserves the style well, but there will still be differences from base versions.


 
 FP8
 
 pruned model stores the UNet and CLIP models in FP8, but the VAE in FP16. This reduces the byte size by almost half in return for a slight quality loss, but most likely won't support hardware-accelerated FP8
 
 inference
 
 (so it won't generate faster).


 
 Tips:
 


 Start with Euler A and a CFG between 5.0 - 9.0 (if using v1.0 then use 4.0 - 7.0). DPM++ samplers can be good too, but CFG range may vary.


 For step count I typically use the formula
 <code>
 5 + cfg * 4
 </code>
 , but 25 is also a good middle ground.