Sign In

聖園ミカ / Mika Misono

88

Download

1 variant available

fp16 SafeTensor

mika_misono_new.safetensors

Half precision, best balance • 288.11 MB

Verified:

Type
LoRA
Stats

1,306

12

3

Reviews
Published

Feb 12, 2023

Base Model

SD 1.5

Training
Steps: 12,000
Epochs: 36
Hash
AutoV2
06D342D695
Trigger Words
mika_misono
default creator card background decoration
Followers - 64

64

Likes - 1182

1.2K

11085-2910943659-mika_misono, ,  (masterpiece), (((extremely detailed CG unity 8k wallpaper))), day, _best quality, illustration, beautiful detai.png

When I tried to find mika's lora, I found that there was no public mika model, so I tried to train one myself. I hope to help some interested friends, and I hope to have a better LORA model to share.

BTW, the angel halo on MIKA's head is included in the training content, but the actual AI often does not leave space for the top of the head, so it is not convenient for Photoshop processing...

For examples (all examples by this):

mika_misono, <lora:mika_misono_new:0.5>, (masterpiece), (((extremely detailed CG unity 8k wallpaper))), day,

best quality, illustration, beautiful detailed girl, lying at the window, depth of field, detailed face and beautiful face,wide-eyed, lace, frills, elegant, muted colors, sweet, sketch by jean - baptiste monge, embroidered, filigreed, long hair, pink hair, white pattern thighhighs, black high heels, yellow eyes,

falling petals, butterfly, mystic atmosphere, pool, after rain, feathers,

sparkle, light rays, lens flare, light particles, hyper detailed, exquisite detail, cinematic lighting, sharp focus, bokeh highlight, cinematic postprocessing,

Negative prompt:

lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts,signature, watermark, username, blurry, artist name,bad feet,mutated hands, (((extralegs)), multiple boys, long back, fewer digits, ((extra legs)), extra arms, mutation

Testing model Anything 4.5 (6e430eb514), even training on Anything 3.0..

Training with 768x512, 512x512, 512x768