[FLUX] PuLID + LORA simple workflow (Basic, Beginner)

Use the [FLUX] PuLID + LORA simple workflow to generate images with consistent faces using PuLID. It takes text prompts and input images, can apply LoRA models for styles or character features, and provides preview and saved-image outputs.

The workflow loads models, encodes text, applies PuLID guidance, samples and decodes an image, and supports comparison, preview, and saving.

Enter a text prompt through CLIPTextEncode and provide an image through LoadImage.

PreviewImage and SaveImage provide preview and saved-image outputs.

The workflow includes step-by-step guidance for connecting and configuring PuLID nodes, plus information about downloading and installing required models.

Model files named in the workflow include: clip_l.safetensors, FLUX_VAE_diffusion_pytorch_model.safetensors, flux1-dev-fp8.safetensors, flux1-dev-Q8_0.gguf, pulid_flux_v0.9.0.safetensors, t5xxl_fp16.safetensors, and t5xxl_fp8_e4m3fn.safetensors.

Named workflow nodes include ApplyPulidFlux, BasicGuider, BasicScheduler, CLIPTextEncode, DualCLIPLoader, DualCLIPLoaderGGUF, EmptySD3LatentImage, FluxGuidance, Image Comparer (rgthree), KSamplerSelect, LoadImage, and Power Lora Loader (rgthree).

Other named workflow nodes are PreviewImage, PulidFluxEvaClipLoader, PulidFluxInsightFaceLoader, PulidFluxModelLoader, RandomNoise, SamplerCustomAdvanced, SaveImage, UNETLoader, UnetLoaderGGUF, VAEDecode, and VAELoader.

The basic node uses a GGUF model, while Options includes ClipLoader and Diffusion Model for switching to standard Flux models.

LoRA models can add specific styles or character features to generated images.

The base model is Flux.1 D.

Before use, verify that pulid_flux_v0.9.0.safetensors, EmptySD3LatentImage, BasicGuider, DualCLIPLoader, and UnetLoaderGGUF are available in your setup.

Vorschlag · nicht geprüft