🎨

Cartoonizing Photos with Comfy Tutorial

Mar 31, 2025

Episode 18: Comfy Tutorial Series - Cartoonizing Photos

Overview

  • Convert photos of people or animals into cartoon characters.
  • Options for SDXL and Flux, including combinations of both.
  • Ensure the latest version is installed.

Getting Started

  • Update Comfy UI:
    • Go to manager and click "Update All" then restart.
  • Interface Settings:
    • Change appearance in settings to match the instructor's.
    • Use new menu options for layout.
    • Set workflow position to top for easier navigation.
    • Badge options: set all to none except source (hide built-in).

Opening Workflows

  • Open workflows from the workflow menu.
  • Multiple workflows can be opened in the same tab.
  • Closing a workflow prompts to save changes.

SDXL Workflow

  1. Upload Image:
    • Start with a portrait photo (e.g., woman, wizard).
  2. Add Short Prompt:
    • Describe the photo briefly.
  3. Run Workflow:
    • Generates a cartoon version of the photo.
    • Not identical, but close enough due to AI's nature.
  4. Experiment with Denoise Value:
    • Lower Denoise = closer to photo.
    • Higher Denoise = more cartoonish.
    • Recommended values for Denoise provided.
  5. Model Download:
    • Download "wildcard XL animation" model from Civit AI.
    • Place in Comfy UI models folder under checkpoints.
  6. Prompt Structure:
    • Short prompt (image description) + long prompt (cartoon style).
    • Use text concatenate node to combine prompts.
  7. Image Processing:
    • Normal image-to-image process using VAE encode.
    • Resize image for optimal results (around 124 pixels).
  8. Control Net:
    • Optionally use pre-processors like Canny or depth map for detail capture.

Testing Animal Images

  • Example: upload a cat image with a simple prompt.
  • Experiment with prompts for different subjects (e.g., dogs, robots).
  • Adjust Denoise and Control Net for desired cartoon effect.

Flux Workflow

  1. Using Laura for Flux:
    • Turbo Alpha Laura for faster generation with eight steps.
    • Split prompt into trigger words and main description.
  2. Scaling Images:
    • Adjust size in multiples of 64 for better results.
  3. Testing Variations:
    • Increase Denoise for more cartoonish effects.
    • Use different Lauras from Civit AI for varied styles.

Hybrid Workflow (SDXL + Flux)

  1. Combine Workflows:
    • Use SDXL workflow on top; Flux workflow below.
  2. Enable and Disable Groups:
    • Fix seed for consistent results across both workflows.
  3. Testing and Tweaking:
    • Adjust prompts for desired features (e.g., red eyes).
  4. Results Comparison:
    • Compare details and effects from both workflows.

Conclusion

  • Join the community for support and updates.
  • Access workflows via Discord channel.
  • Encouragement to like, comment, and subscribe for more content.