A lot of people asked how they could get results like mine using chroma Radiance.
In short you cannot get good results out of the box. You need a good negative prompt like the one I set up and use technical terms in the main prompt like: point lighting, volumetric light, dof, vignette, surface shading, blue and orange colors etc. You don’t neet very long prompts and it tends to lose itself when doing so. It is based on Flux so prompting is closer to flux.
And the most important thing is the wan 2.2 refiner that is also in the workflow. Play around with the denoising, I am using between 0.15 and 0.25 but never ever more, usually 0.20. This also get rids of the grid pattern that is so visible in Chroma radiance and wrong hands and fingers.
The model is very good for “fever dreams” kind of images, abstract, combining materials and elements into something new, playing around with new visual ideas. In a way like SD 1.5 models are.
It is also very hit and miss. While using the same seed allows for tuning the prompt keeping the same rest of the composition and subjects changing the seed radically changes the result so you need to have pacience with it. Imho the results are worth it. Also sometimes you need to correct things in photoshop using generative fill.
The workflow I am using is here .
Here is a small gallery :
