RunDiffusion FX Photorealistic - v1.0

NSFW

RunDiffusion FX Photorealistic

RunDiffusion FX brings ease, versatility, and beautiful image generation to your doorstep.

Join us on our Discord: https://discord.com/invite/wH6dTyBpCf
RunDiffusion.com - Powerful Cloud Servers hosting Automatic1111, InvokeAi and other open source tools.

The RunDiffusion FX Model Series is a game-changer in the realm of image generation. Balancing simplicity and functionality, this model transforms your brief prompts into diverse, imaginative images. Its versatility caters to a spectrum of creative needs, from the novice enthusiast to the seasoned artist. Above all, it's the ability to produce striking outputs that truly sets it apart. See more image examples here.

How to use the model

We set out to create a model that can accept small prompts yet generate amazing results. We've done that with RunDiffusion FX. Watch our guide on YouTube.

Small prompts work great!

(hyperrealism:1.2), BREAK fish swimming under the water BREAK (8K UHD:1.2), (photorealistic:1.2)

Use a Negative!

e.g. Negative Prompt: plain background, boring, plain, standard, homogenous, uncreative, unattractive, opaque, grayscale, monochrome, distorted details, low details, grains, grainy, foggy, dark, blurry, portrait, oversaturated, low contrast, underexposed, overexposed, low-res, low quality, close-up, macro, surreal, multiple views, multiple angles

Wide/Tall aspect ratios are awesome!
- Resolutions like 480x832 832x480 work great then you can upscale.
Lower that CFG! You can use a CFG scale of 3.5 to 5 to get some softer photorealistic images.

Olivio Sarikas Likes it!

We did not ask him to create this video. This was a surprise to us!

License

RunDiffusion supports collaboration and innovation, therefore we are releasing this model free and open for all to use. There are only a few simple rules we'd ask you to adhere to:

Always use the name of the model in any generation service.
- "RunDiffusion FX - Photorealistic"
- "RunDiffusion FX - 2.5D"
Do not change the name or obfuscate it.
Credit us in any merges
Do not resell the model itself
Enjoy

Description

FAQ

Comments (28)

TheHarbingerOfGloryJun 3, 2023

CivitAI

You guys do fantastic work! I'll be sure to recommend you to everyone I can.

GG_SystemJun 3, 2023

CivitAI

it wont let me post a review , civitai is messed up today , ill post a review when it lets me!

dryadlover8Jun 3, 2023

CivitAI

This model is so sick, I can't wait till the site works to post my images

JustMaierJun 3, 2023· 1 reaction

Sorry about that. Should be working now.

xiaozhijasonJun 3, 2023

CivitAI

GG, Great work. Photo level detail.

pursuit_of_beautyJun 3, 2023

CivitAI

Hi, what was this trained on?

It is listed as "trained" not a "merge".

RunDiffusion

Author

Jun 3, 2023· 4 reactions

There's no option in Civitai for "Both". It's base 1.5 SD with concepts added, (people. cars, animals, etc) from here: (and other repos like these)
https://github.com/tobecwb/stable-diffusion-Regularization-Images

https://github.com/JoePenna/Stable-Diffusion-Regularization-Images

https://huggingface.co/datasets/ProGamerGov/StableDiffusion-v1-5-Regularization-Images/tree/main

Then a whole bunch of regularization images from a custom data set curated by a team over 3 months.

601709Jun 3, 2023· 2 reactions

Sus

barryjbatchelor68555Jun 3, 2023

CivitAI

Are you guys going to do an inpaint version?

RunDiffusion

Author

Jun 3, 2023

Obviously if people like this one it’s definitely on the table

ggg__Jun 4, 2023· 1 reaction

@RunDiffusion please make inpainting model, I love this model :)

barryjbatchelor68555Jun 3, 2023· 1 reaction

CivitAI

Does it need a VAE? or was it baked?

RunDiffusion

Author

Jun 3, 2023· 3 reactions

https://huggingface.co/stabilityai/sd-vae-ft-mse-original/blob/main/vae-ft-mse-840000-ema-pruned.ckpt

Use this one

SaffronJun 3, 2023· 2 reactions

CivitAI

Okay....I love this model! It loves wallpaper resolutions! The results are fresh and exciting no matter what I throw at it. Wonderful work!

edconcomic478Jun 3, 2023

CivitAI

Download isn't working at the moment, tries to download "88158.json" then gives a server error, really looking forward to trying these models when I can though

photodarrenJun 7, 2023

Happened to me too. Gave it a second go a few moments later and it worked. Might be server overload or something. Anyway, with luck you tried again and it worked.

onmyballsJun 4, 2023· 2 reactions

CivitAI

I really love the ability to create larger images natively without hires fix. No additional faces or limbs at 704x1024 so far!

gsgsdgJun 4, 2023· 2 reactions

CivitAI

1000% sure you used a non-native text encoder! Great choice, looks neat and I will try it out very soon!

nigelthecorestor7554Jun 4, 2023

CivitAI

I can't seem to get the same image quality on my local install. Am I missing something?

0l1v1aR0551Jun 4, 2023· 1 reaction

I have an amazing experience with this model! The solution is in the promtcraft = read the description of the model, use their examples, simple prompts + negatives ...

xtraJun 4, 2023

Could be your GPU. My ipad cant produce nearly what my laptop can at 512-40step standard, and my laptop isn't anywhere near the e-gpus in quality either, it depends on what you've got powering it

ggg__Jun 4, 2023

as far as i know, different gpu give different result even with same prompt and settings, also using xformers will make the generated images non-deterministic, sorry if i'm wrong

SerenityXJun 7, 2023

@666dxtr6666 I didn't know that! For me, getting output three times as quickly is worth it if I lose consistency...

ggg__Jun 7, 2023· 1 reaction

@SerenityX i'm not an expert (still new with stable diffusion) and not sure about xformers tho, just found from reddit/youtube comment, someone said xformers give different result when enabled even with same settings (not about the quality but the consistency when used same settings), but from the a1111 wiki has mention about --opt-sdp-attention and --opt-sdp-no-mem-attention argument about deterministic & non-deterministic (faster than xformers)

https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Optimizations

also there is a1111 setting that you can change to use either gpu or cpu for "random number generator source" to generate the random seed, set the random seed generator to gpu make the result different depends on the gpu brand,

vae also affect the image quality, so i'd tried changed different vae when the quality is not good/wierd ecspecially using lora & lycoris

and there is weird & wild theory that i have read somewhere that the heat from the hardware affect the quality ;)

ggg__Jun 7, 2023

maybe you can try change the vae or sampler if there are any change, i tried that already and it does change the quality with the same settings

ggg__Jun 7, 2023· 1 reaction

@SerenityX yeah i also enabled xformers because i can generate larger images and good quality :)

AeychpgeJun 9, 2023· 4 reactions

@xtra sorry but...you're using your ipad to generate? h-how?

mariodianJun 12, 2023· 1 reaction

@Aeychpge Draw Things. You can also use it to generate images on your iPhone. It takes quite a while but it's possible.

Checkpoint

SD 1.5