Below is a list of keywords used during the training of this model:
[8386] style
[3471] 1boy
[3167] solo
[2725] blonde hair
[2677] 1girl
[2676] male focus
[2374] no humans
[2154] jacket
[1917] short hair
[1849] sky
[1644] multiple boys
[1642] outdoors
[1352] hat
[1272] night
[1190] 2boys
[1052] english text
[991] blue eyes
[960] black hair
[946] long sleeves
[919] pants
[918] scenery
[900] smile
[866] shirt
[786] tree
[751] cloud
[655] open mouth
[621] indoors
[593] long hair
[585] day
[566] facial hair
[551] brown hair
[546] gloves
[443] standing
[426] upper body
[389] sitting
[373] dark skin
[336] holding
[336] multiple girls
[317] mustache
[287] looking at viewer
[277] weapon
[265] nature
[251] beard
[233] formal
[203] closed eyes
[192] robot
[152] pokemon (creature)
[152] necktie
[132] science fiction
[108] teeth
[91] jewelry
[50] black eyes
[38] earrings
Description
FAQ
Comments (3)
It's interesting to compare the results with what I get from my model made off smaller, curated, dataset and corrected tags (and probably worse video quality). Some of the finer parts are enhanced here, but a lot of things just become the main character's version.
Which tagger do you use? My current one kinda sucks (WD14 moat tagger v2)
@Synthdark8 On-site trainer's tagger+decent number of corrections and filling out some sparesely tagged images on my own.
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.


