GREAT SCOTT! Vivid insanity is here!
===========================================================================
KR34M - 11/25 - Let's give FLUX/KREA the DR34M makeover. This checkpoint excels at photorealism and abstract artwork. We hope you enjoy what's likely our last FLUX release!
Note: NF4 is actually K4_S in GGUF format. FP8 is actually Q8_0, also GGUF!
We recommend the BF16 variant + T5 BF16 if you have the VRAM.
Note: Use LoRA with C4PACITOR at slightly lower than normal weights for best results and to avoid anatomy blending.
===========================================================================
C4PACITOR models are created by DR34MSC4PE with all the trappings you've come to expect as well as some cool bonuses:
Enhanced realism and photo-realistic concepts, trained on high quality datasets with the latest techniques.
Specific anime/illustration tuning to introduce some lost artistic concepts
NSFW tuned and capable of realism and artistic/anime images. Female anatomy is well represented with additional fine tuning in the works.
Exceptional performance with character/other/stacking Lora
Like our work? Buy us a coffee: https://ko-fi.com/dr34msc4pe
===========================================================================
DR34MSC4PE is
c0ur4ge (training/qa/inference code) /
eraser851 (training/captions/tooling code/data/qa)
===========================================================================
Recommendations by base model:
dev - (d_v1/d_v2/a_v)
===========================================================================
CFG: 3-7
Steps: 22-60
Sampler: RES_2M
Scheduler: Beta/Beta57/Bong Tangent
========================================
Hyper8/L16HT - (x-series)
===========================================================================
CFG: 3-6
Steps: 16-32
Sampler: DEIS
========================================
schnell - (s_v0/s_v1):
===========================================================================
Note: 100% Schnell Base
Steps: 2-14
Sampler: DEIS
===========================================================================
Description
We've gone through several test batches trying to get v2's dataset right - this is a fun outcome of the v2 alpha base with some additional enhancements. Thanks for your patience - looking forward to sharing v2!
FAQ
Comments (4)
Thanks for this latest update. I've posted some images to show the versatility of your model, along with prompts and seeds for anyone who'd like to see them. All my images were generated in Forge. Some were made with LoRAs and some without. I hope other Forge users find them useful.
FYI, I use either DEIS or Flux Realistic as samplers most of the time, with Beta, Normal, or KL Optimal as schedulers. Distilled Guidance in my images ranges from 3-7, hovering around 4-5 most of the time. As I get close to the image I want, I change this in tenths and even hundredths to home in on the best setting.
Most LoRAs seem to have been trained on Euler, and I find they need to be turned down to keep from distorting images being rendered in DEIS. 1.0 is much too strong for many of them.
Oh, and on a BTW to c0ur4ge, thanks for the likes and buzz tips and other encouragement. If you check the comments in my newest Dia de los Muertos post, you'll see why the earlier one was replaced. Cheers.
You're very welcome! Thank you for the extra information and kind words. More to come!
I'm currently using both the alpha and beta v2 models. Very versatile and flexible models which fit well with my sometimes eclectic style.
Really looking forward to future versions.
Thank you so much for releasing them and for all the work involved.
You're very welcome - definitely have a few things in the works - I'll keep you posted!
Details
Files
Available On (2 platforms)
Same model published on other platforms. May have additional downloads or version variants.