Qwen-Image-Edit-2511 Lineart Interpolation
About this LoRA
This is a LoRA weight for lineart interpolation, trained on randomly selected 10% of the train subset of Mixamo 240 dataset.
Categories
Examples
Introduction
Qwen-Image-Edit-2511 Lineart Interpolation is a LoRA designed to generate smooth in-between frames from lineart. It's trained to help create fluid transitions between two given lineart images.
This LoRA was trained on a randomly selected 10% subset of the Mixamo 240 dataset.
How to Use
Trigger Word
No trigger word is required.
Recommended Settings
- Base Model: Use with
Qwen/Qwen-Image-Edit-2511. - LoRA Strength: The example code uses a strength of
1.0. You can adjust this value to control the intensity of the interpolation effect. - Prompt: Use
<inbetween> middle frame. The author advises against changing this prompt. - Negative Prompt: The example uses an empty string (
" "). - Inference Steps: The provided Python script uses
30steps.
Python Script
The primary way to use this LoRA is through the provided Python script, which leverages the diffusers library. You'll need to specify the paths to your starting and ending lineart images (CONTROL1_IMAGE_PATH and CONTROL2_IMAGE_PATH) within the script.
FAQ
Q: What's the trigger word?
A: No trigger word is required for this LoRA.
Q: What LoRA strength should I use?
A: The example code sets the LoRA strength to 1.0. Feel free to experiment with this value to get your desired interpolation effect.
Q: What is this model for?
A: This model is specifically for lineart interpolation. It takes two lineart images and generates an "in-between" frame, useful for animation or smooth transitions.
Q: Is this model ready for production?
A: The author notes it's still in an "attempt phase" and hasn't been fully verified, though it works well in some examples. Keep this in mind for critical projects.
Q: What prompt should I use?
A: You should use the prompt <inbetween> middle frame. The author recommends not changing it.
Technical Details
- Base Model:
Qwen/Qwen-Image-Edit-2511 - Training Info:
- Trained for 3000 steps.
- Used 10% of the training subset from the Mixamo 240 dataset.
- Source: Hugging Face
