Sign In

Angkor Wat (Angkor Ruins)

20

160

271

11

Verified:

SafeTensor

Type

LoRA

Stats

160

271

96

Reviews

Published

Aug 13, 2024

Base Model

Pony

Training

Steps: 432
Epochs: 10

Usage Tips

Clip Skip: 1
Strength: 1

Trigger Words

Angkol_ruins

Hash

AutoV2
943404BFD4

Created on Civitai

One cool thing you could do in SD1.5 (which I discovered when trying to make backgrounds for Golden Sun characters) was to add “Angkor Ruins” to a prompt’s background tags: It would combine well with almost anything and make a neat fantasy vibe. Unfortunately XL and PXL don’t really have this ability to do this, or even recognize the keyword for its actual use. I doubt this will get nearly as good as 1.5 could pull by default, but I want to try. Please post the pics you make (even failures): Seeing people use my LoRA without me getting to see what they make is sad, and every unique user that posts an image gets me buzz to train future LoRAs (it’s 10 users to break even). Effect is best when paired with a style LoRA (I don't know why).

Trigger: Angkol_ruins (sic)
Supporting tags: Doorway, hallway, window, arch, pillar, column, engraving, stone stairs, stone wall, overgrown (for trees and other flora growing within the stone), ruins (for the general weathered appearance), rubble (for actual piles of rubble), outdoors, building, scenery (for outdoors), stone wall, stone floor, indoors (indoors), moss (set a lower strength if you don't want your scene looking like Vault 22), stone head, statue (for the carved stone faces. Just statue for normal kind. Beware most checkpoints will try to make the faces very... alive, especially if moss or foliage is in the prompt),
Neg: Photo (Medium), door (creates the thing with a knob on hinges)

Training data was 54 photos that were Public Domain, CC-0 and CC-By (see bottom for credits. Wish Civitai supported putting this in a txt file) images off Wikimedia Commons. Auto-tagging was generally cooperative, except for thinking everything was water (and thus a bridge), so tagging went pretty smoothly. May do v2 with more training data if needed, but hopefully won’t have to.

CC-By credits:
Modification: These were obviously used to train the resource. I don't think they were knowingly modified beyond resizing and compression.
https://commons.wikimedia.org/wiki/File:Angkor-079_hg.jpg by H. Grobe
https://commons.wikimedia.org/wiki/File:Angkor_Wat_(6043006666).jpg by Christine Zenino from Chicago, US
https://commons.wikimedia.org/wiki/File:Angkor_Wat_-_070_Cloister_(8580576947).jpg by Photo Dharma from Penang, Malaysia
https://commons.wikimedia.org/wiki/File:Angkor_Wat_Tempel_06.jpg by Olaf Tausch
https://commons.wikimedia.org/wiki/File:Second_level_-_Angkor_Wat_(6202414540).jpg by Jorge Láscar from Australia
https://commons.wikimedia.org/wiki/File:One_of_the_interlinked_galleries_of_the_second_level_-_Angkor_Wat_(6202408484).jpg by Jorge Láscar from Australia
https://commons.wikimedia.org/wiki/File:The_shadow_(21805220259).jpg by cloud.shepherd
https://commons.wikimedia.org/wiki/File:Angkor_Thom_interior_2009view.jpg by Allie Caulfield
https://commons.wikimedia.org/wiki/File:2009-09-03_09-07_Siem_Reap_130_Angkor,_Thommanon.jpg by Allie Caulfield
https://commons.wikimedia.org/wiki/File:Bayon_Temple_Angkorwat_(58)_(27952358742).jpg by Sasha India
https://commons.wikimedia.org/wiki/File:Bayon_Temple_Angkorwat_(56)_(27952359122).jpg by Sasha India
https://commons.wikimedia.org/wiki/File:Bayon_Temple_Angkorwat_(60)_(27977375881).jpg by Sasha India
https://commons.wikimedia.org/wiki/File:Bayon_Temple_Angkorwat_(57)_(28054714925).jpg by Sasha India
https://commons.wikimedia.org/wiki/File:Bayon_Temple_Angkorwat_(31)_(27441126944).jpg by Sasha India
https://commons.wikimedia.org/wiki/File:Bayon_0757_(27773814510).jpg by Sasha India
https://commons.wikimedia.org/wiki/File:Bayon_0756_(27773814860).jpg by Sasha India
https://commons.wikimedia.org/wiki/File:Prasat_Kravan_0780_(28020157786).jpg by Sasha India
https://commons.wikimedia.org/wiki/File:Stay_Angkor_Wat_(149828123).jpeg by Zz Lighting