An experimental hypernetwork trained on 100+ different anime characters. Do we actually need to create specific networks for each and every character or could we work on improving our base models instead?
Model source
Source description
This continues training of the same hypernetwork and adds a bunch of new characters. The pre-existing characters are improved while the new one have varying quality.
hayasaka ai
noire (neptune series)
rebecca (cyberpunk) - eyes not quite working
shibuya rin
sonozaki mion
tedeza rize
sawamura spencer eriri
c.c.
rory mercury
Sources
1 sourceModel artifacts
1 artifactSource excerpts
2 excerptsSource context: 1254 downloads · Type Hypernetwork · Base model Other
alisa reinford
emma millstein
fie claussell
sara valestein
akame (akame ga kill!)
esdeath
horikita suzune
ichinose honami (youjitsu)
karuizawa kei
kushida kikyou
ayanokouji kiyotaka
shizuka hanekura - replaces 'akatsuki (log horizon)'
shiroe (log horizon)
shokuhou misaki
misaka mikoto
shirai kuroko
saten ruiko
kamijou touma
aiz wallenstein
high elf archer (goblin slayer!)
These characters were added even later and some certainly requires more training to be useful:
koko (kamitsubaki studio)
shiba miyuki
saegusa mayumi
fubuki (one-punch man)
tatsumaki
kitagawa marin
inui sajuna
kamado nezuko
aizawa tomo
yuzuriha inori
sinon
tohru (maidragon)
aurora sya lis kaymin
mumei (kabaneri)
fischl (genshin impact)
The following tags were changed to avoid them having partial overlap with other characters, this caused shiroe to look like akatsuki for example:
karyl_(princess_connect!) -> karyl
elaina_(majo_no_tabitabi) -> elaina
keqing_(genshin_impact) -> keqing
lumine_(genshin_impact) -> lumine
hu_tao_(genshin_impact -> hu_tao
eula_(genshin_impact) -> eula
Source context: 1229 downloads · Type Hypernetwork · Base model Other