CashMoney v1.0E + Niji
A rework of the original checkpoint!
Changes:
More vibrant and consistent
Better details and backgrounds
Fixed VAE
Niji version is more highly stylised
Previous updates:
CashMoney v5.0
This is the biggest update to the model yet.
Make it your own with over 4,000 trained artist style hints (create your own looks by combining artist hints!)
Character and style support not available in base Pony models.
Works with Pony and SDXL LoRAs.
No need for any aesthetic or quality (e.g pony_9) tags in positive prompt, save your tokens with simplified prompting. Simple negative prompt of lowres, worst quality, low quality, bad anatomy, bad hands works well.
Trained using powerful 10 x A40 GPUs over 2 million steps.
Released without EA paywall to thank everyone for your support, likes and downloads.
CashMoney v4.0
This is the best version of this model and I’m really happy to upload it. Improvements:
Better prompt adherence
Clearer, more vibrant gens
The model is better at lewd things so I just went with it for the preview images
Even better RNG (more likely to produce a good image in a given generation)
The same TE as v1.0 for Lora compatibility
Great anatomy and quite dynamic
Responds well to different CFG scales
I am making this model early access for the following reason:
I blew a tidy sum of money and a whole lot of time messing around with cloud GPUs to improve the model. It isn’t yet the best it can be, but I’ve learned some hard lessons I’ll be putting forward to future versions and other models. Those of you kind enough to drop $1 of buzz for the model are in a special place in my heart! I know Early Access is a pain but it’s great for model creators who somehow need to sustain their passion/hobbies/obsessions.
UPDATE: v3.5 should resolve Pony Lora issues.
UPDATE: Added v3 which uses feedback from v2. It’s got more of the personality of v1, while keeping improvements of v2. Also doesn’t explode with some Pony LoRAs. I think…
UPDATE: If your favourite Pony LoRAs don’t work well with this model try using my Pony LoRA Adapter.
UPDATE: You can now use this model directly on the CivitAI generator service.
Low RNG
Vibrant and dynamic gens
Detailed backgrounds, but also respects simple/plain background prompt
SFW and NSFW gens (use rating_explicit Pony tag)
Dynamic poses and lighting
Anime and illustration style
Not only waifus
Description
Better sharpness, details and anatomy
FAQ
Comments (7)
Nice Job Man :)
top tier checkpoint but is it possible to have v1 available on web generator also?
Please make v1 an option to select on web generators 😢
Sadly V3 absolutely butchers any loras even with adapter lora :( V1 is still one of my alltime fav models
I do agree with a lot of the posts here. V1 I do believe was the best version of this model so far. V2 messed up Lora’s pretty bad and made it nearly incompatible with most to all Lora’s within Pony, however v1 didn’t have this issue at all, was clean, refined, and polished in itself. V3 just introduced more and more issues, I believe taking away even more from the unique style of V1. Now, I may be biased due to the fact I loved v1 model. But that is simply just my personal opinion on this model.
The model no longer shows up in search and there is also no create option on the page anymore. Will it be back soon?