Meet Splashed Flux — the long-awaited fusion of Splashed Mix’s signature photorealism with Flux’s deep, prompt-aware intelligence. It delivers that clean, realistic output you know and love, now with even sharper brains behind the beauty. Insanely quick renders, built on the same ethos as the main XL range. Easily out performs base Flux on speed, athestics and ability to produce beautful and creative images.
Rendering at a base of 768x1024 takes as little as 28 seconds in my tests, using 10 steps and an upscale. Speed and beauty, like a Bond villian.
Recommended settings:
– 8–10 steps
– Euler (simple), DPM++ 2M (simple) sampler
– 3.5 - 5 guidance scale
Technical notes:
– Single merged checkpoint with CLIP + T5 + VAE
– Plug-and-play with ComfyUI, Forge, and Swarm — just load the checkpoint and you're good to go
Every image you see has been chosen to show just how far this model can go — from varied art styles and bold creative directions, to being the greatest AI photographer you'll ever use. We’ve pushed it to its limit, proving that fewer steps doesn’t mean lower quality. “What if Flux thought and looked like Splashed?”
Description
Say hello to Splashed Flux — the long-awaited fusion of Splashed Mix’s signature photorealism and Flux’s powerful, prompt-aware brain. Built for creators who want gorgeous, accurate images without the trade-offs, this is our answer to: “What if Flux thought like Splashed?”
Run it with:
10–15 steps for best results
Euler (simple) sampler
3.5 guidance scale
Expect stunning realism, rich texture, and refined edge control — all with the intuitive feel of Splashed and the deep interpretive power of Flux.
FAQ
Comments (8)
"Splashed node" missing in ComfyUI
Missing Node Types - workflow>Flux_Simple
This model is the fastest Flux.1 D I used so far. Image quality is top notch. Skin looks in most cases very natural. No need to use any Skin/Shine Loras. It follows prompts very well.
This is a great Flux model for prompt spamming, with 11 second renders, the only thing slowing it down is my imagination.
The recommend FG of 3.5 is perfect for the majority of prompts, but I would recommend increasing the FG to achieve your desired outcome when it comes to the more out there prompts.
Thanks to Splashed for yet another awesome model.
Thank you very much! I'm new to this, and before, generating on Flux was easier (just by downloading a checkpoint). But now I'm faced with a problem - not a single Flux was working. I got on you and following your instructions, I got a generation! Thank you so much!
Probably one of the best models based on Flux, or even the best. Yes, there can sometimes be artifacts during generation, but the generation speed compensates for this completely. And the results are simply amazing. It works well with loras, not as perfectly as the basic Flux, but something like 4/5. Anyway, the result images are just great and detailed. Thank you very much to Splashed!
need a GGUF of this :(
Will this model suffer from increased step count if we can handle the time it takes? (i.e. having an rtx 5090) or is it just that it can be coherent at 15 steps.
waiting for V2.0
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.