Flux.2 Workflow with optional Multi-image reference

Use this Flux.2 workflow to create images from prompts, with optional reference images and LoRA, and choose initial, upscaled, or blended outputs; it loads the model, CLIP, and VAE.

Start by loading the model, CLIP, VAE, and optional LoRA, then write a prompt that can be split into edit instructions and overall text.

Use the sampler with common sampling options, plus optional upscaling and blending.

Provide a prompt, optionally split into separate edit-instruction and overall text.

Choose whether to use reference images; four are included, and more can be added by copying a reference-image node.

Output options include an initial image, an upscaled image, or a blend of the two; normal-and-upscaled batch handling saves both the initial and upscaled images.

This workflow uses Flux.2 D and is published by Yorgash.

The workflow lists these node packs: ComfyUI-Easy-Use, ComfyUI-Crystools, ComfyUI-KJNodes, and ComfyUI-utils-nodes.

Listed model files are flux2-dev-fp8_scaled.safetensors, flux2-vae.safetensors, mistral_3_small_flux2_fp8.safetensors, aidmaMJ61Flux.2v0.5.safetensors, and RealESRGAN_x4plus_anime_6B.pth.

Turn reference images on or off, split the prompt into edit instructions and overall text, and choose normal, upscaled, or blended output handling.

The listed node-pack versions are ComfyUI-Crystools 1.27.3, ComfyUI-Easy-Use 1.3.3, ComfyUI-KJNodes 1.1.7, and ComfyUI-utils-nodes 1.3.7.

Before setup, verify the exact files for flux2-dev-fp8_scaled.safetensors, flux2-vae.safetensors, and aidmaMJ61Flux.2v0.5.safetensors; the supplied listing gives no further identifying details for these files.

Vorschlag · nicht geprüft

The listing does not specify a license.