360Redmond LoRA: 360° Panorama Images in Stable Diffusion XL
Create 360° view and panorama images with 360Redmond, an SDXL LoRA by ArtificialGuyBR. Trigger words, recommended settings, and tips.
On this page
360Redmond LoRA: Generate 360° View and Panorama Images with SDXL
TL;DR:
- 360Redmond is a LoRA fine-tuned on Stable Diffusion XL 1.0 that generates 360° view and panorama images in a wide variety of themes.
- Trigger it with the instance prompt
360, 360 view. - Generate at 1600x800 and upscale 3x–4x; keep prompts simple and view results in a 360° panorama viewer.
What 360Redmond does
360Redmond is a versatile LoRA for Stable Diffusion XL that pushes generation toward 360° view and panorama output. According to the model card, it produces 360 view and panorama images in "a wide variety of themes," which makes it a general-purpose concept LoRA rather than a single-subject one. The output is meant to be viewed with an online 360° panorama viewer (the card links 360° Panorama Viewer Online by renderstuff.com), so the images behave like equirectangular panorama frames rather than ordinary landscape crops. It fits alongside the other SDXL LoRAs by ArtificialGuyBR.
How it was trained
The author, ArtificialGuyBR, notes that GPU time from Redmond.AI allowed the training run to finish. The LoRA is fine-tuned on Stability AI's stable-diffusion-xl-base-1.0, so it needs an SDXL workflow in ComfyUI, Automatic1111, or any diffusers-based pipeline. The model card does not publish dataset size, training steps, or learning rate, so nothing more can be said about the training procedure beyond the base model and the acknowledgment.
Trigger words
The instance prompt on the model card is:
360, 360 view
The author recommends including 360 or 360 view in prompts "to get better 360" results. Keep the rest of the prompt simple: the card explicitly warns that complex scenes — for example office interiors or dense forests — are where the LoRA "ends up making mistakes." Simple prompts and uncluttered scenes generate the most reliable panoramas.
How to use it
Any SDXL LoRA loader will do:
- ComfyUI: load the LoRA with an SDXL base checkpoint, then prompt with
360 viewplus your subject. This follows the standard ComfyUI SDXL workflow. - Automatic1111: drop the safetensors file in
models/Lora, select it, and add, 360 viewto the positive prompt. - diffusers: load with
pipe.load_lora_weights("artificialguybr/360Redmond")and prompt with360 view.
Exact LoRA strength values are not published in the model card — start with the defaults and adjust up or down to taste.
Recommended settings
The model card gives two concrete tips:
- Generate at 1600x800, then upscale 3x or 4x for better quality. This pairs naturally with an SDXL upscaling pass.
- Don't expect a perfect result on the first pass — the author says the LoRA "sometimes needs more than one gen to create good images."
Common errors and fixes
Error | Cause | Fix | Source
|---|---|---|---
Distorted or messy output in complex scenes | LoRA struggles with dense compositions such as office interiors and forests | Simplify the prompt; fewer subjects and plain backgrounds | Model card Panorama looks wrong in a normal image viewer | 360° output is designed for panorama viewers, not flat viewing | Open the image in a 360° panorama viewer | Model card Soft or low-detail results | Generating at default SDXL resolution | Use 1600x800 and upscale 3x–4x | Model card No panorama effect | Trigger words missing from prompt | Add 360 or 360 view | Model card
FAQ
What base model does 360Redmond need? SDXL 1.0 (stabilityai/stable-diffusion-xl-base-1.0).
What license is it released under? bigscience-openrail-m.
What resolution should I generate at? The card recommends 1600x800, then 3x–4x upscaling.
Does it work with any theme? The card describes it as versatile across "a wide variety of themes," with the caveat that complex scenes produce mistakes.
Will every generation be good? No — the author says the LoRA sometimes needs more than one generation.