QWEN-EDIT_360

A Qwen Image Edit LoRA for generating 360° equirectangular HDRI panoramas from a single input image.

The base Qwen Image model struggles to produce usable equirectangular output. This LoRA was trained specifically to fix that — turning any input frame into a full 360° scene that can be used as an HDRI, projected into 3D, or wrapped around a virtual camera for AI filmmaking workflows.

GitHub: github.com/mickmumpitz/QWEN-EDIT_360 — ComfyUI workflow and documentation.

Usage

Place the LoRA in your ComfyUI models/loras/ folder and load it on top of Qwen Image Edit.

Prompt

Create a 360 panoramic image of [very short description of surrounding scene].
An HDRI image captured on a 360 degree camera, insta 360, equirectangular
projection. Keep the style the same.

ComfyUI workflow

A workflow is provided with these key inputs:

• Input image — any standard image you want to extend into a 360° scene
• Padding — adjustable around the input frame
• Seed — change it if the result isn't satisfying

Seamless output

The workflow produces seamless 360° images by automatically shifting the seam to the middle of the image and inpainting it. Tune the mask size, mask blur, and seed if you run into artifacts at the wrap point.

Training

• Base model: Qwen/Qwen-Image-Edit
• Dataset: 20 equirectangular HDRI image pairs sourced from Polyhaven (CC0)

Results aren't perfect yet, but they're significantly better than the base model on equirectangular output.

License

MIT. Training data is CC0 (Polyhaven).

Author

Created by mickmumpitz.