VR-360-Outpaint-LTX2.3-IC-LoRA Morphs Clips Into 360 VR Scenes

A delicate slender glass widescreen video monitor hovers with no bezel showing a ghostly city at golden hour.

The VR-360-Outpaint-LTX2.3-IC-LoRA is a proof-of-concept adapter that turns standard widescreen video clips into full 360-degree equirectangular footage for VR viewing. It operates as an IC-LoRA on top of the LTX-2.3 model, filling in masked areas of a projected cinemascope frame to create an immersive sphere. The release bundles sample outputs, a ready-to-use ComfyUI workflow, and the LoRA weights themselves.

TheBurgstall developed this early v0.1 project to test the feasibility of outpainting 2.39:1 video into a 360° format. Training was deliberately constrained to a small curated set of semi-static city scenes and run on a single H100 GPU. The narrow scope helped validate the approach before future, more diverse expansions.

Turning widescreen clips into 360-degree immersive video

Key Features
  • Outpaints 2.39:1 clips to 360° equirectangular video.
  • IC-LoRA adapter for the LTX-2.3 pipeline.
  • Includes ComfyUI workflow and side-by-side samples.
  • Trigger word “equirectangular” helps guide content.
  • Tuned for semi-static city scenes at ~100° FOV.
  • Outputs viewable in any VR or 360 player.

This release is for creators who want to test 360 outpainting locally without relying on cloud services. It suits people comfortable with ComfyUI who can work within its narrow sweet spot of static urban shots. Early adopters can use it to explore the technique and prepare for the more robust versions that are planned.

What’s next and current rough edges

TheBurgstall plans a larger-scale training run covering interiors, landscapes, and varied camera motion. For now, the v0.1 model struggles with fast cuts, the top and bottom polar caps, and complex prompts—it’s not a replacement for native 360° capture. Training was done on one H100 across 3,500 steps, resulting in a 1.3 GB LoRA file.

“This is an early v0.1 release. Expect rough edges, limited subject variety, and inconsistent coherence outside the sweet spot described below.” — Source: Hugging Face