jbilcke-hf/flux-dev-panorama-lora-2

Prompt

HDRI panoramic view of TOK, visiting an amusement park about harry potter.

Prompt

HDRI panoramic view of TOK, inside of a beautiful japanese onsen with sakura tree, hot baths

Prompt

HDRI panoramic view of TOK, inside of an art gallery, very modern, luxurious stylish, with expensive marble and copper-plated stuff

Prompt

HDRI panoramic view of TOK, inside of a fighter jet cockpit, flying in the air, over an aircraft carrier, in San Francisco bay

FLUX.1-[dev] Panorama LoRA (v2)

A LoRA model to generate panoramas using Flux dev.

Which image ratio to use?

This model has been trained on images with a 2:1 ratio (2048x1024).

So you might get good results is you use this for width and height.

However panorama viewers are pretty flexible when it comes to resolution, and FLUX.1 seems to generalize well.

For instance the gallery samples have been generated in 1536 × 640 (~21:9), since this is reasonably fast (16 sec on the HF Inference API).

It doesn't work for case X or Y

It usually work fine for "normal" requests, but the model might have trouble creating the spherical distortion if you ask for uncommon content, locations or angles.

If you give me some examples, maybe I can try to find more data to better cover uncommon panoramas.

Non-commercial use

As the base model is FLUX.1-[dev] and since the data comes from Google Street View, it should be used for non-commercial, personal or demonstration purposes only.

Please use it responsibly, thank you!