Glamour - CivArchive (CivitAI Archive)

Glamour - SDXL v1.0

This model adds a touch of gloss and luxury dream to any scene. Use it with a weight of 0.1 to 1. It doesn't require a trigger word, but it pairs nicely with glamour-style, glamour, and gloss. When applied to an inherently glamorous scene, your mind will swirl with all the glitter and success.

SD Version

It was trained at 1024×1024. SD can have issues with hires, so I do recommend using Hires fix from 768×768 or 512×512 at Denoising strength 0.5-0.7 (f.i. with a latent upscaler)

The project in SD+SDXL is a continuously evolving work in progress...

Description

endep. test tag
{
  "unetLR": 0.0005,
  "clipSkip": 1,
  "loraType": "lora",
  "keepTokens": 0,
  "networkDim": 32,
  "numRepeats": 2,
  "resolution": 1024,
  "lrScheduler": "cosine_with_restarts",
  "minSnrGamma": 5,
  "noiseOffset": 0.1,
  "targetSteps": 3666,
  "enableBucket": true,
  "networkAlpha": 16,
  "optimizerType": "Adafactor",
  "textEncoderLR": 0.00005,
  "maxTrainEpochs": 13,
  "shuffleCaption": false,
  "trainBatchSize": 4,
  "flipAugmentation": true,
  "lrSchedulerNumCycles": 3
}

LORA

SDXL 1.0

by airesearch

Download (Beta) View on CivitAI