Sign In

Living Basalt, Stilt Canals & Geode Architecture (SDXL)

0

Not recently loaded

Download

1 variant available

fp16 SafeTensor

jsolarum_architecture_xl_v1.safetensors

Half precision, best balance • 162.65 MB

Verified:

Type
LoRA
Stats

22

0

Reviews
Published

Sep 7, 2026

Base Model
Hash
AutoV2
49AB27BE49
Trigger Words
jsolarum_architecture
04_arch_01_living_basalt_sanctuary.png

Field test · Gallery Run 01 (2026-10-05)

Trigger: jsolarum_architecture.

Seven prompts it wasn't trained for, at 0.75, trigger first: https://civitai.com/posts/31472449

The G2 prompt (the masked figure on the velvet stage): https://civitai.com/posts/31472796

Stacked with other LoRAs at 0.5: https://civitai.com/posts/31482206 · https://civitai.com/posts/31482223

How the run was done, and every post: https://civitai.com/articles/36214/i-put-64-of-my-loras-on-the-same-eight-prompts-heres-what-each-one-actually-does

> Solarum Architecture XL renders the monumental built environments, habitat modules, and ancient sanctuaries across the system.

>

> - Living Basalt Sanctuaries: Massive cyclopean masonry grown rather than carved, interlocking basalt monoliths bound by bioluminescent fungal grout, and towering skylit colonnades.

> - Isla Buena Canal Residences: Multi-tiered timber and woven-bamboo stilt compounds elevated above azure tidal channels, connected by rope walkways and lantern-lit bridges.

> - Oort Reef Geode Cities: Hollowed interior nickel-iron asteroid caverns lined with terrace dwellings, central zero-g hanging gardens, and spinning atmospheric ring habs.

> - Enceladus Undercity Corridors: Pressurized ice-tunnels reinforced with brass ribbing, steam condensation vents, and glowing cyan lichen conduits.

> - Training Details: 246 images, 1,600 steps (about 26 passes per image), trained on Juggernaut XL v9 at 768 px.

- Trigger Words: jsolarum_architecture

- Recommended Strength: 0.5–1.0, default 0.65. With long, descriptive prompts the look is already there at 0.5–0.75; short prompts (like "a train station concourse") come through better nearer 1.0. Above 1.0 it starts drawing made-up signage and pulls scenes toward the same look (lanterns, clocks, round skylights). It also nudges people toward one look, so keep it lower in scenes with characters. Tested on Juggernaut XL v9.

Thanks to DetroitWobbly, whose CEF evaluation #0004 of this LoRA prompted the retest behind the strength range above.