Guess you are all familiar with @Lodestone's Chroma models based on FLUX.1-schnell: <a target="_blank" rel="ugc" href="https://civarchive.com/models/1330309/chroma">https://civarchive.com/models/1330309/chroma</a> Chroma1-HD works in latent space (compressed representation) like traditional AI image generators.But this is about: <a rel="ugc" href="https://huggingface.co/lodestones/Chroma1-Radiance">Chroma1-Radiance</a> - a radically tuned version that works directly in pixel space, skipping the latent-to-image decode step entirely.The main advantages of Radiance:<ul><li>Sharper output with more accurate colors</li><li>Reduced artifacts and distortion</li><li>Eliminates color artifacts and blurring that can occur when decoding from latent space</li></ul>In simple terms: HD uses the traditional compressed method (latent space → decode → image), while Radiance generates images directly in pixel space, resulting in higher quality with fewer artifacts.You'll need a diffusion model file. I did use this one: <a target="_blank" rel="ugc" href="https://huggingface.co/lodestones/Chroma1-Radiance/blob/main/latest_x0_full_20M_dataset_run_1024.safetensors">https://huggingface.co/lodestones/Chroma1-Radiance/blob/main/latest_x0_full_20M_dataset_run_1024.safetensors</a>And a text encoder. I've been using the same as for Chroma-HD: <a target="_blank" rel="ugc" href="https://huggingface.co/silveroxides/flan-t5-xxl-encoder-only/resolve/main/flan-t5-xxl-fp16.safetensors">https://huggingface.co/silveroxides/flan-t5-xxl-encoder-only/resolve/main/flan-t5-xxl-fp16.safetensors</a>No need for a VAE.Just pull the showcase images into Comfy. Workflow is embedded.

Files with SHA256: fbb40208514991adcb9d58cb3a2011d5b386593e27143e30574cb96d36c6d71b