Qwen Image Edit 2509 Caricature LoRA

Model description

Qwen Image Edit 2509 Caricature LoRA

📌 Description

This LoRA transforms input images into sketched caricature art with exaggerated features. It's an image-to-image model that takes your photo as input and creates humorous, artistic caricature representations of people and animal subjects with emphasized facial features and characteristics.

📝 Usage

Use the following prompts:

create a caricature of this image
create a caricature of these people
create a caricature with exaggerated funny features of this image

Trigger word: caricature

For best results, include the trigger word in your prompt to activate the caricature style transformation.

🚀 Best Results

This model pairs well with the following LoRAs for faster and more refined results:

Qwen Image Edit 2509 Lightning 4-step
Qwen Image Edit 2509 Lightning 8-step

⚡ These LoRAs are optional and only improve speed.
The base model works fine without them.

📦 Available Checkpoints

This model includes 8 training checkpoints to give you control over the creative style:

1000 steps
2000 steps
2250 steps ✨ Good choice (Creative style)
2500 steps ⭐ Top recommended checkpoint
2750 steps ✨ Good choice (balanced style)
3000 steps
3500 steps

🎨 Checkpoint Selection Guide

Lower steps (1000-2250): More creative and experimental caricature style
Balanced (2500): Best overall balance ⭐ Top choice
Alternative balanced (2250/2750): Good alternatives with slightly different characteristics ✨
Higher steps (3000-3500): More standardized illustration style with less creative variation

💡 Tip: Start with checkpoint 2500 for the best results, then experiment with 2250 for more creativity or 2750-3500 for a more standardized illustrated look.

🔄 Troubleshooting

If the generated caricature doesn't show enough exaggeration, try adding "exaggerated funny features" to your prompt
Change the seed if the result doesn't match your expectations
Ensure the input image has clear facial features for best caricature results
Try different checkpoints if the style isn't what you're looking for

🛠️ Training Information

Training Pairs: 220
Data Type: Synthetic dataset
Input: Photos of people/animals/subjects
Output: Sketched caricature representations with exaggerated features
Style: Hand-drawn sketch aesthetic with comedic proportions

Download model

Download them in the Files & versions tab.