ComfyUI Basic Operations

Part 3: Your Quick Guide to ComfyUI's Image-to-Image Workflow

Willow

• 2 min read
Part 3: Your Quick Guide to ComfyUI's Image-to-Image Workflow
This image was created using ComfyUI on the Cephalon AI platform.

Already grasp text-to-image in ComfyUI? Great! Image-to-image is your next logical step. It builds on that foundation by letting you use an existing image as a starting point. This guide cuts to the chase, showing you the key nodes and how to wire them up to get started fast.

The Key Nodes You'll Need

Load Image

  • What it does: Uploads your starter image.
  • Find it: Right-click -> Add Node -> image -> Load Image.

VAE Encode

  • What it does: Translates your uploaded image into the AI's internal language ("latent space").
  • Find it: Right-click -> Add Node -> latent -> VAE Encode.
  • Remember: This is the Encoder (not the Decoder).

K Sampler

  • What it does: The core engine. It uses your image and prompts to create a new image in the latent space.
  • Key Tip: The output size matches your uploaded image, so skip the "Empty Latent Image" node.

VAE Decode

  • What it does: Translates the AI's internal data back into a visible image.
  • Note: It usually just uses the VAE from your model.

Preview Image

  • What it does: Shows you the final result.

How to Connect It All

Follow these steps to build your workflow:

Hook Up the Model:

From your Checkpoint Loader, connect:

  • model -> K Sampler's model
  • CLIP -> both CLIP Text Encode nodes (rename them to "positive" and "negative").
  • VAE -> both VAE Encode and VAE Decode nodes' VAE input.

Write Your Prompts:

  • Type what you want into the "positive" node.
  • Type what you want to avoid into the "negative" node.
  • Connect the condition output of each to the K Sampler's positive and negative inputs.
  • Don't forget: Your words are just as important as the image for guiding the AI.

Load and Encode Your Image:

  • Upload your picture in the Load Image node.
  • Connect its image -> VAE Encode's image.
  • Connect VAE Encode's latent -> K Sampler's latent.
  • This tells the sampler what to use and what size to make the new image.

Decode and See:

  • Connect K Sampler's latent -> VAE Decode's latent.
  • Connect VAE Decode's image -> Preview Image's image.

You're all set! Just adjust your settings in the K Sampler (steps, CFG, etc.), hit "Queue Prompt", and watch your new image generate.

Unlock Full-Powered AI Creation!
Experience ComfyUI online instantly:👉 https://market.cephalon.ai/aigc
Join our global creator community:👉 https://discord.gg/KeRrXtDfjt
Collaborate with creators worldwide & get real-time admin support.

Sign up for more like this.

Enter your email
Subscribe