LoRA Part 2: Text-to-Image Practice with LoRA — An Example of Practical Case

In the LoRA Part 1, we expanded on the article in the Part 2 by introducing how to integrate the LoRA model node into the text-to-image workflow to create a new workflow. In this part, we'll provide a practical example of a workflow that includes the LoRA model node.

Practical Example: Generating a Chinese-style anime girl

Now, let’s fill in the parameters and generate an actual image.

Step 1: Select a Model

l  In the Checkpoint Loader node, choose a model suitable for anime style, such as meinamix_meinaV11.safetensors.

l  In the LoRA Loader node, choose a model suitable for Chinese-style illustration type, such as 国风插画SDXL.safetensors，set the strength_model to 0.7, and keep the strength_clip unchanged.

Step 2: Enter Prompts

• Positive Prompt: Enter the content you want to see, e.g.:
      masterpiece, best quality, (hanfu), 1girl, white plum blossom on one side, 1 girl, solo, front view, close up
• Negative Prompt: Enter low-quality content to avoid, e.g.:
      lowres, text, error, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg

💡 Tip: Use clear, well-lit images for the best results.

Step 3: Set Dimensions

In the Empty Latent node, set the width to 512, height to 768, and batch size to 1.

Step 4: Configure the Sampler

Set the key parameters in the K Sampler node:

Random Seed: 0

After Running: Random

Steps: 25

CFG: 7

Sampler: dpmpp_2m

Scheduler: karras

Denoise: 1.0

Step 5: Generate and Save the Image

Click “Add Prompt Queue” to start generating.

In the Preview Image node, right-click and choose “Save Image”.

Results and Notes

With this workflow, you can successfully convert text into exquisite image. Keep in mind:

Size: Output matches the input resolution.

Clarity: Results may be slightly blurry.

Details: Expressions and poses may change a little.

These issues can be improved later with upscaling workflows or ControlNet.

Suggestions

Experiment: Different photos may require different prompts or parameters.

Practice: Try several runs to find the best settings.

Save: Keep successful workflows for reuse.

Unlock Full-Powered AI Creation!
Experience ComfyUI online instantly:
https://market.cephalon.ai/share/register-landing?invite_id=RS3EwW
Join our global creator community:
https://discord.gg/MSEkCDfNSW
Collaborate with creators worldwide & get real-time admin support.