Sign In

sdprompt - lora to enhance __sdprompt__ wildcard trained with no data (animated gifs)

63

524

12.8k

4

Updated: Oct 5, 2024

toolpopularconceptmod

Verified:

SafeTensor

Type

LoRA

Stats

524

12.8k

7

Reviews

Published

May 25, 2023

Base Model

SD 1.5

Training

Steps: 2,000
Epochs: 1

Trigger Words

__sdprompt__

Hash

AutoV2
8C5B21DF29
Supporter Badge May 2023
ntc's Avatar

ntc

Trained for 2000 steps on:
art detailed concept painting digital highly sharp trending intricate focus illustration lighting portrait smooth elegant cinematic beautiful fantasy render realistic face style high hair body wearing eyes matte light character dramatic black details hyper dark ultra shot background detail epic volumetric girl photorealistic symmetrical cyberpunk blue masterpiece oil long 8k red female studio composition glowing hd vibrant quality artwork award man cute design fine dynamic shaded perfect painted resolution winning soft--

using https://github.com/ntc-ai/conceptmod

and reversed with --save_opposite

The idea is to fill up the token limit with the most commonly repeating words (minus artist and site names)

Use with https://civitai.com/models/69651?modelVersionId=74311

animation loras from -1 to 1 with __sdprompt__ (https://civitai.com/models/69651?modelVersionId=74311)