Workflow Overview

This workflow is designed for poster face swapping and online image generation using Stable Diffusion XL and InstantID. It reconstructs facial features while preserving the original poster style. The process involves ControlNet for structure preservation, face analysis (InstantIDFaceAnalysis), and color matching (ColorMatch) to create a seamless swap.

Core Models

1. Stable Diffusion XL (DreamShaper XL v2.1 Turbo)

   • Generates the base image.

2. InstantID (ip-adapter_instant_id_sdxl)

   • Captures facial features and ensures facial consistency.

3. ControlNet (control_instant_id_sdxl & xinsir_controlnet-canny-sdxl_V2)

   • Preserves structural elements such as outlines and key facial features.

4. SUPIR (SUPIR-v0F_fp16.safetensors)

   • Performs super-resolution upscaling for higher-quality output.

Key Components (Nodes)

1. CheckpointLoaderSimple (Loads SDXL Model)

   • Loads DreamShaper XL v2.1 Turbo.

2. InstantIDModelLoader (Loads InstantID)

   • Analyzes facial details for precise swapping.

3. ControlNetApplyAdvanced (Applies ControlNet)

   • Ensures structural integrity post face swap.

4. ImageResize+ (Resizes Images)

   • Adjusts resolution to match the target.

5. ApplyInstantID (Executes Face Swap)

   • Transfers facial identity while preserving style.

6. VAEDecode (Decodes Latent Space)

   • Converts latent representations into images.

7. SUPIR_model_loader_v2 (Loads Super-Resolution Model)

   • Enhances final output quality.

8. ColorMatch (Performs Color Correction)

   • Ensures consistent skin tone.

Workflow Structure

This workflow consists of four major stages:

1. Input & Preprocessing

• Loads the original poster image.

• Extracts facial details via InstantIDFaceAnalysis.

• Uses LineArtPreprocessor to maintain structure.

2. Face Swapping & Image Generation

• InstantID + ControlNet refine face alignment.

• KSampler handles diffusion-based face generation.

3. Refinement & Optimization

• SUPIR Super-Resolution enhances details.

• MaskBlur+ and CropMask refine the face mask.

4. Final Output

• ColorMatch corrects skin tone.

• SaveImage stores the result.

Input & Output

Inputs

• Original poster image

• Target face image

• ControlNet sketch

• Text prompts

• Face blending intensity

Outputs

• High-resolution swapped poster image

• Ready-to-save or further edit

Considerations

1. Performance

   • Requires RTX 3090 or higher for smooth operation.

2. Troubleshooting

   • Adjust InstantID strength if swap is inaccurate.

   • Use ColorMatch to fix skin tone mismatches.

3. Compatibility

   • Works with ComfyUI 1.3+ and requires some custom nodes.

This analysis provides a detailed guide for beginners and intermediate users to understand and utilize this face-swapping workflow in ComfyUI! 🚀