🛠️ Workflow Overview

Purpose and Function:
This workflow automatically converts real-life photos into anime-style images. It combines upscaling models, VAE encoding/decoding, sampling, and facial enhancement to generate high-quality anime images with refined facial features and improved resolution.

Core Features:

1. Image Preprocessing: Loads and resizes the input image.

2. Anime Style Conversion: Uses custom anime-style models to transform the real image.

3. Facial Refinement: Enhances facial details and fixes imperfections.

4. Final Output: Displays and saves the high-resolution anime-style image.

🔥 Core Models

1. WAI_NSFW-illustrious-SDXL_v11

   • Function:

     • The primary model for transforming real-life photos into anime-style images.

   • Installation:

     • Install using ComfyUI Manager or manually:

     • Download the .safetensors file and place it in the models/Stable-diffusion directory.

2. 4x-AnimeSharp (Upscaling Model)

   • Function:

     • Enhances the image resolution and sharpens details.

   • Installation:

     • Place the model file in models/UpscaleModels.

     • Alternatively, install it using ComfyUI Manager.

⚙️ Nodes Explanation

1. LoadImage

   • Function:

     • Loads the real-life image into the workflow.

   • Input:

     • Image file path.

   • Output:

     • Image data.

2. UpscaleModelLoader

   • Function:

     • Loads the upscaling model used to enhance the image resolution.

   • Parameters:

     • 4x-AnimeSharp

   • Output:

     • Upscale model.

3. ImageUpscaleWithModel

   • Function:

     • Applies the upscaling model to enlarge the image.

   • Input:

     • Image

     • Upscale model

   • Output:

     • Upscaled image.

4. VAEEncode

   • Function:

     • Encodes the image into the latent space for further processing.

   • Input:

     • Image

     • VAE model

   • Output:

     • Latent image data.

5. KSampler

   • Function:

     • Samples the latent image to generate the anime-style output.

   • Parameters:

     • Sampling method: euler_ancestral

     • Sampling steps: 30

     • CFG scale: 0.6

   • Input:

     • Model

     • Positive and negative prompts

     • Latent image

   • Output:

     • Latent anime-style image.

6. VAEDecode

   • Function:

     • Decodes the latent image into a visual image.

   • Input:

     • Latent image

     • VAE model

   • Output:

     • Anime-style image.

7. CLIPTextEncode

   • Function:

     • Encodes the textual prompt into conditioning data.

   • Input:

     • Text prompt.

   • Output:

     • Conditioning data (CONDITIONING).

8. CLIPSetLastLayer

   • Function:

     • Adjusts the last layer of the CLIP model for better prompt guidance.

   • Input:

     • CLIP model.

   • Output:

     • Modified CLIP model.

9. SaveImage

   • Function:

     • Saves the final anime-style image to the local storage.

   • Input:

     • Image.

   • Output:

     • None (saves the file).

🧩 Workflow Structure

✅ Group 1: Image Upload

• LoadImage → Loads the real-life image.

• UpscaleModelLoader → Loads the 4x-AnimeSharp model.

• ImageUpscaleWithModel → Applies the upscaling model to enhance image resolution.

✅ Group 2: Model Prompts

• CLIPSetLastLayer → Adjusts the CLIP model's final layer.

• CLIPTextEncode → Applies positive and negative prompts:

• Positive prompts:

• masterpiece, best quality, amazing quality

• Negative prompts:

• teeth, cleavage, (worst quality:1.65), (low quality:1.2), low resolution, watermark, dark spots, blemishes, dull eyes, wrong teeth, red teeth, bad tooth, multiple people, broken eyelashes

✅ Group 3: Initial Image Processing

• VAEEncode → Encodes the image into the latent space.

• KSampler → Samples the latent image using the anime model.

• VAEDecode → Decodes the sampled latent image into the final anime-style image.

✅ Group 4: Facial Refinement

• The workflow uses facial enhancement to fix imperfections, such as blurry or distorted facial features.

✅ Group 5: 4K Upscaling

• scale → Applies 4K upscaling to improve image quality and resolution.

✅ Group 6: Final Output

• SaveImage → Saves the final anime-style image to the local storage.

🔥 Inputs & Outputs

✅ Inputs:

• Real-life image.

• CLIP positive and negative prompts.

• Upscaling model.

• Anime-style generation model.

• Sampling parameters (steps, CFG scale, etc.).

✅ Outputs:

• High-resolution anime-style image with refined facial details.

⚠️ Considerations

1. Hardware Requirements:

   • This workflow involves multiple model loads, VAE encoding/decoding, and upscaling, requiring a GPU with at least 12GB of VRAM for optimal performance.

2. Resolution Limitations:

   • High-resolution input images (above 1600x1200) may cause VRAM overflow.

   • To avoid instability, limit the image size to 1600x1200 or smaller.

3. Compatibility:

   • Ensure the ComfyUI and model versions are compatible to prevent errors.

4. Output Quality Control:

   • Use negative prompts to filter unwanted artifacts, such as noise, blurriness, or facial imperfections.