ArtificialGuyBR

Home / Blog / Training LoRAs

360 Panoramic Flux LoRA Guide

How to use the 360View Redmond Flux Klein 9B LoRA to generate panoramic cityscapes, landscapes, and coastal scenes in FLUX.2.

2 sources cited Training LoRAs

360 Panoramic Flux LoRA Guide

A compact FLUX 2 Klein 9B LoRA for generating 360-degree panoramic scenes.

What it does

This LoRA adapts Black Forest Labs FLUX.2-klein-9B toward wide panoramic compositions. Instead of standard framed subjects, it pushes outputs into the 2:1 equirectangular-style aspect ratio used for immersive environments, cityscape concept art, and wide landscape references.

Panoramic example output 1

Panoramic example output 2

Panoramic example output 3

The 2:1 output ratio is consistent across the gallery samples, confirming the model targets wide-field, panoramic framing rather than standard viewport compositions.

Training and base model

The adapter targets black-forest-labs/FLUX.2-klein-9B, a smaller FLUX 2 variant. Training was supported by Redmond.AI compute. The LoRA is released under the Apache 2.0 license.

How to use

  1. Load the base model FLUX.2-klein-9B.
  2. Attach the LoRA with standard diffusers or ComfyUI LoRA loading nodes.
  3. Include 360 VIEW or 360. in the prompt.
  4. Use a panoramic resolution, e.g. 1920x960, 2048x1024, or wider.
  5. Keep CFG and sampler settings consistent with FLUX 2 defaults; no special scheduler is required.

Prompt style still matters. Pair the trigger with scene keywords: night city, mountain ridge, or tropical beach.

Common errors and fixes

Error | Cause | Fix | Source

|---|---|---|---

Narrow crop instead of panorama | Missing trigger or low LoRA weight | Add 360 VIEW and increase LoRA weight slightly | HF card Distorted horizon | Extreme aspect ratio mismatch | Start at 1920x960 or 2048x1024 | Output samples Missing detail in sky/ground | Over-compressed generation | Use FLUX 2 default steps and CFG range | HF card

FAQ

Does this work in A1111? Yes, because it is a standard LoRA. The author recommends ComfyUI, but Automatic1111 can load it via the LoRA tab.

Can I use it for VR 360 images? It produces wide panoramic stills, but not cubemap or stereoscopic 360 rig outputs. Treat it as environment concept art, not final equirectangular source.

What triggers should I use? The card lists 360 VIEW and 360. as the main triggers. Adding scene keywords keeps outputs coherent.

Sources