Join The Tinkerer on Whop to get early releases, private pages, and the Tinkerer Discord role - all in one place. š Join on Whop
After a ton of requests, Iām finally rolling out CyberRealistic Flux (FLUX.1 dev)! Itās designed to make realistic images, both safe-for-work and not-so-safe-for-work. Itās not perfect yet, but itās a solid start and sets things up for whatās coming next.
Heads up: The (non AIO release) doesnāt include clip models or VAE. Youāll need to grab those separately here.
(clip_lorg.safetensors = orginal clip_l file. clip_l.safetensors is the CLIP-SAE-ViT-L-14 one)
Settings I Use:
Sampling method: DPM++ 2M / Euler
Sampling steps: 20/30 Steps
Distilled CFG Scale: 2ā3.5 (Samples were done at 3.5, but lower values often give better results)ā ļø A Quick Warning:
This model can make mature or sensitive stuff. Youāre in charge of what you create, so use it wisely (or at least get creative with it).
š¬ Want Better Prompts?
Thereās a custom ChatGPT thatāll whip up great prompts for CyberRealistic Flux:
š [Try it now on ChatGPT]
Description
In this version, Iāve made improvements to skin texture and overall clarity - itās much cleaner now. The model is fully NSFW-capable, so donāt worry š
Iāve included a range of different examples, all created without any LoRAs. There will only be a few NSFW samples, but it should be clear that this model can definitely handle them.
Get it via Early Paid Access on Civitai or through Patreon (Discord download).
FAQ
Comments (44)
Can this work on Stable diffusion webui, or not ?
It can work on Forge/Forge Neo - not on A111.
@CyberdeliaĀ Thank you, Sir
Cyberdelia, I'm still your biggest fan. I tell you this because maybe you were worried I'm not these days, but i'm still the biggest one nonetheless. No question! So don't worry about it.
About the checkpoint, i don't use Flux, but i'm sure is great if it was made by you.
Take care.
Why did you switch to the "Early Paid" practice?
Itās just my sneaky little way of getting people to join my Patreon. Gotta do what you gotta do, right?
Because @Cyberdelia deserves to be rewarded for hard work. Weird concept right?
@GasparGamesĀ Absolutely. And I'd give you an award for your sycophancy.
Life is expensive man.
Hello i buy the version and im very happy with it. but do you have a workflow for it maybe it works with yours better. and im not on patreon and i dont want to go to patreon. So can you help me with it? Greetings and thx for this beautiful model:)
Workflow as in ComfyUI? I don't use it - I'm a die-hard Forge Neo user :)
@CyberdeliaĀ ah so its the normal flux workflow?:)
thx for the fast anwser. And by the way you make so good work thats my first diffusion model for that i payed so please keep on :)
@CyberdeliaĀ I use the old Forge. Is Neo better for flux and new models? Old forge is extra slow with flux. It doesnt fully offload the encoders and vae to system ram, like Comfy. My 8gb Vram card can handle only Nf4 with acceptable speed.
@hermeswisam568Ā Iām not entirely sure - I use three different versions of Forge (donāt ask š ). But yes, Iād recommend upgrading to the latest Forge just to stay current. That said, keep in mind that more features can sometimes mean things run a bit slower. If you have the space, just install it next to old Forge and use 1 location for models, lora's etc.
@CyberdeliaĀ for me it works awesome with the flux dev setting and some fine tuning :)
@CyberdeliaĀ I know you said "don't ask"......but I have to ask ^^ lol. As a Forge main user myself mostly; I'm curious for a rundown to all the pros & cons of each fork of it?
I'm using such a heavily customised personal version myself, that it gives me headaches even thinking of swapping sometimes lol. But at this point, the thing is starting to feel like it's held together with duct tape haha.
@ShadowCellĀ The new version supports WAN and Qwen, and honestly, if you donāt need those, I donāt really see a reason to switch. Every version has different settings or components that have been removed or added. It can also happen that some plugins stop working (Iāve converted a few myself, but that requires some programming knowledge). So that can be a pro but mosty it's a con.
So itās pretty clear what my answer to your question would be: should you switch? No. If it works fine for you, just keep using it. Iām running three different versions myself, and sometimes I realize Iāve been using the original Forge instead of Forge Neo without even noticing. Itās not like the newest version creates much better images - it just supports some newer features and can be a bit faster. But is that really worth losing your settings and custom tweaks? I donāt think so.
I am unable to install version 2.5 with Invoke AI.
Invoke AI always crashes when I attempt to install it.
Would it work with an AIO version of CyberRealistic Flux 2.5?
Are both the launcher and Invoke itself up to date (or as close to the latest version as possible) on your setup?
@CyberdeliaĀ Yes, I have installed InvokeAI version v6.9.0rc3.
I use the launcher for model installation.
This was displayed after the crash in the console: Hashing cyberrealisticFlux_v25.safetensors: 100%|##########| 1/1 [04:01<00:00, 241.56s/file]Hashing cyberrealisticFlux_v25.safetensors: 100%|## ########| 1/1 [04:01<00:00, 241.57s/file]
Process exited with code 3221225477
@mwsynch322Ā Ah, what you see is a error hashing the file. Try to re-download the file again.
When Invoke starts up or loads a model, it hashes .safetensors files to verify integrity and identify them.
If the file was incompletely downloaded (cut off mid-way, or corrupted during transfer), the hasher tries to read past the valid range of bytes ā native library crash
@CyberdeliaĀ Please excuse me for only getting back to you now.
I have found the error.
However, I am still not sure why it only occurs with this model.
I had installed Invoke AI on an external hard drive, and after moving the installation to the C partition, I was able to install your model.
I am looking forward to seeing the results of the first test with the CyberRealistic Flux 2.5 fp8 model.
However, it will take some time as I only have a GTX 1650 graphics card.
Translated with DeepL.com (free version)
Excellent model. I've had to really load it down with loras to produce any Cronenbergian monsters. Can't even get extra fingers or toes out of it without asking for them. 10/10
Which resolutions are best to use with this? I'm assuming 1024x1024, but other flux models have different resolutions.
Flux.1-dev is natively 1408x1408, but other fine tunes like Krea and Chroma are best at 1536x1536 and 1024x1024 respectively, so I'm just curious what the best resolutions are for your fine tune are?
I suggest genning at 1920x1088 if your GPU can handle it, or equivalent lower resolutions as long as you're going over 1MP.
i loved it, however on comfyUi, only the v2.1 works, when I try to load the 2.5 it just crashes without any specific error, with the same exact settings
your cyberpony model is great with penises, but this one seems quite bad. any chance you'll make that point of focus in future versions?
Dear @Cyberdelia will GGUF version 2.5 be released, or is it not worth waiting for?
Will upload it tomorrow/Wednesday (need to create a few samples).
Thank you very much, very grateful. All your model is a best!
Still planning on releasing a GGUF version of 2.5?
Just discovered this model and I'm really loving it. Can't wait for the GGUF 2.5!
@supergangĀ @Ratime855Ā @imnotanevilrobotĀ Uploaded... sorry for the long wait!
@CyberdeliaĀ It's like Christmas! Your work is so good, thank you.
This checkpoint is the only Flux one I use, I love the outputs and the quickness of the workflow I had.
With the recent update to Comfyui, it seems the rgthree nodes are no longer compatible. Do you have a workflow that works really well on your checkpoint that doesn't utilize that now outdated node @Cyberdelia ?
@CyberdeliaĀ Thank you for the link. I do find it a bit crass to plug that without any explanation or remote empathy towards the question at hand. Even the slightest response prior to plugging a paid service goes a long way in business. Plugging the link alone more so feels as if you couldn't even be bothered to respond. Thanks for what you've provided, good luck with what you provide in the future
@HawtDayumĀ Well, you could have answered that differently instead of the way you did. I had set several Patreon posts to public and assumed Iād done the same with this one. Turns out I hadnāt, and you made that very clear in your own way. Itās set to free now...
@CyberdeliaĀ So, just to clarify, I voice to you my discontent for something you responded with you "thought" you'd done but hadn't. In turn, you tell me I could have voiced my discontent differently? Please provide me with how you prefer I voice myself towards you so that I may make that correction in the future (not being sarcastic, just trying to move forward amicably).
I have no issue with you or the product you provide for this community. It just came across as you solely wanting to plug the paid product. You didn't intend to so I don't hold it against you. Again, thank you for what you provide and thank you for taking time to clarify the overall intent.
@HawtDayum No idea how you should have responded differently :) I honestly just thought the link was set to public - it was meant as helpful advice and not to plug anythingā¦ Doesnāt matter. Itās sorted now.
@HawtDayumĀ But you can still support me on Patreon. Itās always appreciated, considering all the free work I do for you guys ;)
Good output, easy to use, versatile. Great all around Flux model.
Like always you make a beautiful work
Details
Files
Available On (2 platforms)
Same model published on other platforms. May have additional downloads or version variants.