【FLUX】IMG to IMG — v5.0d

Use 【FLUX】IMG to IMG — v5.0d in ComfyUI to turn a source image and prompt into a generated image with Flux.1 D; you can optionally apply SeedVR2 tiled upscaling and save generation parameters with the image.

Load a model and source image, write a prompt, and generate; the workflow is described as using eight nodes.

The auto version can download models automatically.

Inputs are a source image and prompt; manual setup also lists model, CLIP or text-encoder, and VAE files for the base or GGUF options.

Output images can carry all generation parameters for online publishing and remixing.

In ComfyUI, turn on “Use Nodes V2 (Vue),” refresh, and then load the workflow.

Install ComfyUI-UmeAiRT-Toolkit through ComfyUI Manager by searching “UmeAiRT,” or use the UmeAiRT Auto-Installer.

Base setup files listed: flux1-dev-fp8.safetensors or flux1-dev-fp16.safetensors; clip_l.safetensors; t5xxl_fp8_e4m3fn.safetensors or ViT-L-14-TEXT-detail-improved-hiT-GmP-TE-only-HF.safetensors; and ae.safetensors.

GGUF setup files are labeled Q8, Q6, Q5, Q4, or Q3; named files include flux1-dev-Q8_0.gguf and t5-v1_1-xxl-encoder-Q8_0.gguf. A listed upscaler file is 4x_NMKD-Siax_200k.pth.

Named workflow parts include BasicScheduler; DualCLIPLoaderGGUF; Fast Groups Bypasser (rgthree); Image Comparer (rgthree); KSamplerSelect; LoadImage; mxSlider; Params In // RvTools; RandomNoise; Text Input [Dream]; UNETLoader; UnetLoaderGGUF; UpscaleModelLoader; and VAELoader.

Use the three LoRA slots with individual on/off toggles and strength controls; additional LoRA modules can be connected.

Toggle the built-in SeedVR2 tiled upscaler on or off.

ComfyUI may have display problems if “Use Nodes V2 (Vue)” is not enabled before loading.

Before downloading or loading, verify that the linked page is 【FLUX】IMG to IMG — v5.0d, that your chosen setup has matching model, text-encoder, and VAE files, and inspect the loaded workflow’s node-to-node wiring.

Suggestion · non vérifié