Image & Video

Rafaelldestilo's Chroma V48 DC Workflow delivers optimized Midjourney-style image generation

ComfyUI workflow for Chroma V48 model with 2x/4x upscaling and LoRa support

Deep Dive

Rafaelldestilo has released the Chroma Workflow V48 DC, a tailored ComfyUI setup for the Chroma-unlocked-v48-detail-calibrated model, which delivers Midjourney-style aesthetics. The workflow simplifies aspect ratio selection—1:1, 2:3, and 3:2—with Civitai synchronization for activation keys. It features two upscaling options: Lexica, which adds sharpness and detail, and NMDK, which refines details at 2x or 4x (the 2x version is downscaled from 4x to save space). The workflow also includes a LoRa Manager for optional LoRa loading, though the creator notes the base model is already stunning without LoRAs.

The V48 model was trained at 512x resolution, so the workflow advises against generating directly at 1024x to avoid quality loss; instead, users should generate at native resolution and upscale. Aesthetic mode 10 is enabled by default for optimal results, with an option to disable. The workflow also supports Sage Attention for performance improvements. Rafaelldestilo plans to update by possibly removing the LoRa Manager and will release a Klein 9b workflow and Zimage Turbo soon. This tool is ideal for professionals seeking consistent, high-quality image generation with minimal manual tuning.

Key Points
  • Supports aspect ratios 1:1, 2:3, 3:2 with Civitai synchronization for activation keys.
  • Two upscalers: Lexica (sharpness/detail) and NMDK (2x/4x refinement with space-saving option).
  • Integrated LoRa Manager for optional LoRa support, but base V48 model is recommended without LoRa.

Why It Matters

Streamlined ComfyUI workflow for high-quality Midjourney-style generation, offering professional upscaling and flexibility.