How to use from the
Use from the
Diffusers library
pip install -U diffusers transformers accelerate
import torch
from diffusers import DiffusionPipeline

# switch to "mps" for apple devices
pipe = DiffusionPipeline.from_pretrained("black-forest-labs/FLUX.2-klein-4B", dtype=torch.bfloat16, device_map="cuda")
pipe.load_lora_weights("althjs/sig_f_child_male")

prompt = "hejin_sig_f, close-up portrait of a child boy, casual t-shirt, calm smile, Studio Ghibli watercolor"
image = pipe(prompt).images[0]

sig_f_child_male

5~10์„ธ ๋‚จ์ž ์•„์ด ยท ์–ด๋ฆฐ ์™•์ž / ์–ด๋ฆฐ ์‹œ์ ˆ ์ฃผ์ธ๊ณต

ํ•œ๊ตญ YouTube ์ฝ˜ํ…์ธ  ์ž๋™ํ™” ํŒŒ์ดํ”„๋ผ์ธ hejin ์˜ ์‹œ๊ทธ๋‹ˆ์ฒ˜ ์บ๋ฆญํ„ฐ ๋ผ์ด๋ธŒ๋Ÿฌ๋ฆฌ LoRA.

ํ•ต์‹ฌ ์ •๋ณด

ํ•ญ๋ชฉ ๊ฐ’
Trigger word hejin_sig_f
๊ถŒ์žฅ strength 0.8
Base model Flux 2 Klein 4B
๋ผ์ด์„ผ์Šค apache-2.0
ํŒŒ์ผ๋ช… sig_f_child_male_klein4b_v1.safetensors

์บ๋ฆญํ„ฐ ์ •๋ณด

์™ธ๋ชจ ๋ฒ ์ด์Šค: a young boy around 7 to 10 years old with innocent gentle charm, soft tousled brown or golden hair, large round curious eyes that radiate wonder, smooth pale skin, small slender build, often a soft thoughtful smile or wide-eyed gaze of discovery. ๊ถŒ์žฅ ์˜์ƒ prior: classic prince-like outfit / casual modern boy clothes (t-shirt and shorts) / school uniform / hanbok ์–ด๋ฆฐ์ด / ๋™ํ™” ์† ์˜์ƒ. ๊ถŒ์žฅ ๋ถ„์œ„๊ธฐ: ์–ด๋ฆฐ ์™•์ž, ๋™ํ™” ์ฃผ์ธ๊ณต, ์–ด๋ฆฐ ์‹œ์ ˆ ํšŒ์ƒ ์žฅ๋ฉด, ์ฒœ์ง„๋‚œ๋งŒํ•œ ์•„์ด. ํ™œ์šฉ ์ž‘ํ’ˆ ์˜ˆ: ์–ด๋ฆฐ ์™•์ž (Le Petit Prince), ๋™ํ™” ์‹œ๋ฆฌ์ฆˆ, ์ฃผ์ธ๊ณต์˜ ์–ด๋ฆฐ ์‹œ์ ˆ flashback. ํ•™์Šต ์‹œ caption โ€” ๋‹ค์–‘ํ•œ ์˜์ƒ/๋ฐฐ๊ฒฝ/ํ‘œ์ •. dataset ์ž‘์„ฑ prompt ๋ฒ ์ด์Šค: 'a young boy around 8 years old with soft tousled blonde hair, large innocent curious eyes, gentle thoughtful expression, painterly Studio Ghibli watercolor, dreamy soft lighting.'

ํ•™์Šต ์ •๋ณด

  • ๋ฐ์ดํ„ฐ์…‹: 20 ์ปท (1024ร—1024) โ€” ์˜์ƒ / ํ‘œ์ • / ๋ฐฐ๊ฒฝ ๋‹ค์–‘ํ™”
  • Caption ํ˜•์‹: hejin_sig_f, <framing> of <identity>, <์˜์ƒ 1~2 ๋‹จ์–ด>, <ํ‘œ์ • 1~2 ๋‹จ์–ด> (60~90 chars sweet spot)
  • Caption ์›์น™: face ๋ฌ˜์‚ฌ (๋ˆˆ์ƒ‰ยท๋จธ๋ฆฌ ๊ธธ์ดยทํ”ผ๋ถ€ ๋“ฑ) ๋Š” caption ์— ๋ฐ•์ง€ ์•Š๊ณ  trigger word ๋งŒ์œผ๋กœ face ํ˜ธ์ถœ โ€” face ํ•™์Šต ์‹ ํ˜ธ dominant
  • Base model: Flux 2 Klein 4B
  • ํ•™์Šต ๋„๊ตฌ: ai-toolkit (Ostris) โ€” rank 24 / alpha 16 / lr 1e-4 / steps 1500 / flowmatch scheduler

์‚ฌ์šฉ ์˜ˆ์‹œ โ€” WaveSpeed Cloud API

curl -X POST https://api.wavespeed.ai/api/v3/wavespeed-ai/flux-2-klein-4b/text-to-image-lora \
  -H "Authorization: Bearer $WAVESPEED_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "prompt": "hejin_sig_f, close-up portrait of a child boy, casual t-shirt, calm smile, Studio Ghibli watercolor",
    "size": "1024*1024",
    "seed": 12345,
    "loras": [{"path": "https://huggingface.co/althjs/sig_f_child_male/resolve/main/sig_f_child_male_klein4b_v1.safetensors", "scale": 0.8}]
  }'

๋ผ์ด์„ผ์Šค

์ด LoRA ๋Š” base model Flux 2 Klein 4B ์˜ ๋ผ์ด์„ผ์Šค (apache-2.0) ๋ฅผ ๋”ฐ๋ฅธ๋‹ค.

Apache 2.0 โ€” ์ƒ์—… ์‚ฌ์šฉ / ์ˆ˜์ • / ์žฌ๋ฐฐํฌ ์ž์œ . royalty-free.

ํ•™์Šต ๋ฐ์ดํ„ฐ์…‹ ๋ฏธ๋ฆฌ๋ณด๊ธฐ

ํ•™์Šต์— ์‚ฌ์šฉ๋œ dataset ์ผ๋ถ€ (์ „์ฒด 20์ปท์€ dataset/ ํด๋” ์ฐธ์กฐ):

01 05 10 15
#01 #05 #10 #15

๊ฐ ์ปท์˜ caption ํŒŒ์ผ (dataset/NN.txt) ์€ ํ•™์Šต ์‹œ ์‚ฌ์šฉ๋œ ์งง์€ ํ˜•์‹ (60~90 chars) ์œผ๋กœ trigger + framing + identity + ์˜์ƒ + ํ‘œ์ •๋งŒ ํฌํ•จ.

Downloads last month
11
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support

Model tree for althjs/sig_f_child_male

Adapter
(53)
this model