Create AI Images : Image to image

 Image generated using the Z-image Turbo model (image to image) πŸ”»

Image generated using the Z-image Turbo model (image to image)


In my previous posts, we explored how to utilize Text-to-Image (txt2img) on lower-end PC setups and how various parameters influence the final output. Today, we’re shifting our focus to a more dynamic tool: Image-to-Image (img2img).

Using the Z-Image-Turbo (ZIT) model once again, I’ll show you how to achieve stunning results without needing a high-end workstation. While results vary based on individual settings, this guide focuses on generating images that stay faithful to the original composition while enhancing quality and style.



Workflow & LoRA Setup

To follow along, you’ll need the custom workflow and the specific LoRA model I’ve prepared. (Note: A LoRA(Low Rank Adaptation) is a specialized tool that allows the AI to generate specific styles—like pixel art or oil painting—with high efficiency and precision.)

  1. Download: Access the files via Download Workflow & LoRA Here.

  2. Installation: Place the downloaded LoRA file (pixel_art_style_z_image_turbo.safetensors) into the ComfyUI/models/loras directory.

  3. Launch: Open ComfyUI and press Ctrl + O to load the workflow(json) file.



The Workflow: Transforming Your Images

Once the workflow is loaded, you will see the following layout:

layout

  • Step 1: Locate the Load Image node (highlighted in the red box) and upload the image you wish to transform. For this demonstration, I used an image of Shorekeeper from Wuthering Waves.

layout step1

  • Step 2: Enter your prompt. For this demonstration, I used a prompt designed to maintain a realistic aesthetic without drifting too far from the original photo.

step2

  • Step 3: Queue the prompt. After a brief processing period, you’ll see the transformation.

step3

 The result is surprisingly crisp and aesthetically pleasing.



Controlling Creativity via Denoising & CFG

The true power of img2img lies in fine-tuning your parameters. While you can experiment with many settings, two are crucial for ZIT:

  • CFG (Classifier Free Guidance): For this model, I recommend keeping this between 1.0 and 2.0. It keeps the generation stable and aligned with the prompt.

  • Denoising Strength: This is your primary lever for "creativity." Think of Denoising as "how much of the original image should the AI erase and redraw?"

  1. Low Denoising (0.3): The AI stays very close to the original structure, making subtle refinements.

    Low Denoising

  2. High Denoising (0.9): The AI treats the original image as a loose sketch, generating something almost entirely new.

    High Denoising


Wrapping

Today, we’ve only scratched the surface of what’s possible with img2img. By customizing your workflows or switching models, the possibilities are endless—from forcing specific poses to maintaining character consistency across different actions, or even training your own custom models. 

In the next post, we will dive deeper into Pose Control and Custom Model Training

Comments

Popular posts from this blog

Training Custom LoRA for Z-Image-Turbo (ZIT) via Local PC

Create AI Images : High-Speed Pose Control & Custom Character LoRA Workflow for ZiT

Creating AI Videos on an RTX 4060 (LTX-2 vs. Wan2)