"PNTE" Negative Stable Diffusion LoRA
Increase the quality and amount of details in images with these negative LoRAs for Stable Diffusion models.
How to use:
THESE ARE MEANT TO BE USED WITH NEGATIVE STRENGTH VALUES.
To do this simply add the LoRA to the negative prompt or manually adjust the LoRA strength depending on the diffusion interface you are using.
For ease of use, ComfyUI users can use the SnakeOil custom_node suite I created. This node not only automatically inverts negative LoRAs but it also looks for them in models/nloras folder rather than the normal models/loras folder. This helps with organization and keeping our ever growing lists of LoRAs a little shorter.
SDXL:
A version my SD2.1 "point-e" negative embedding for use with XL.
Keep CLIP strength at 1.0 and adjust the strength of the LoRA to preference.
LoRA Strength can range from -0.01 to -2.00.
Description
Use negative strength value to add details and coherence to your images.
FAQ
Comments (10)
Interesting, but I'm quite unclear on what this does, more information and examples would be helpful.
Will do, it is meant to be used with negative strength to improve overall image results.
this is amazing, thank you
Can you tell me what approach you used to train this LoRA? It's not a slider by any chance?
Anyway, I would be very grateful if you could share the repository and a few words about specifics.
This was trained on the seed images for COCO CLIP R-Precision evaluations from Open-AI's point-e.
Will it have versions for Flux.1 Schnell and Flux.1 Dev? I like this lora.
Dev and Schnell are both cut down versions of Flux that do not support negative LoRA values so unless they let me train Flux Pro no chance of any Flux version sadly.
@doctor_diffusion please pay attention to de-destilled Flux versions. Many users give feedback that they get better training results on this model, versus distilled Flux Dev
@NotBadImBread I have not explored the de-destilled version of Flux yet. Thank you for mentioning this, I will take a look and see what is out there and if this technique will work.
@NotBadImBread Have a working De-distilled Flux Negative LoRA that you cant quite push past -0.25 but it works to that point before getting strange, will share soon but I have such a back log of models to release.
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.