Files

T

Antigravity Agent f8b21af4be Wiki cleanup: error-doc removal, dedup merge, link normalization

10_Wiki/Topics 대규모 정리:
- 오류 캡처/미완성 stub 문서 227개 제거
- 교차폴더 중복 43클러스터 병합 (63파일 → redirect)
- 링크명 정규화: 깨진 링크 수정·redirect 직결·개념 매핑 ~2,400건
- 카테고리 MOC 6개 신규 생성
- Graph 섹션 미해결 related-keyword 링크 10,058건 제거

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

2026-05-20 23:52:15 +09:00

9.3 KiB

Raw Blame History

id, title, category, status, canonical_id, aliases, duplicate_of, source_trust_level, confidence_score, verification_status, tags, raw_sources, last_reinforced, github_commit, tech_stack

title

Commercial AI Art Production

매 한 줄

"매 ad / mockup / brand asset 의 AI 의 production". 매 prompt engineering + 매 reference + 매 draft mode + 매 license-aware. 매 modern: 매 Midjourney V7 draft, 매 Adobe Firefly (commercial-safe), 매 Flux 의 photoreal. 매 typography 의 still 의 manual.

매 핵심

매 commercial use case

Product mockup: 매 e-commerce.
Ad creative: 매 social, banner.
Brand asset: 매 hero image, banner.
Social media: 매 daily post.
Logo / icon.
Storyboard / pre-vis.
Editorial illustration.
Game asset.

매 platform 별 strength

Platform	Strength	Weakness
Midjourney	매 aesthetic, mood	Text, exact spec
DALL-E 3	Text in image, follow instruction	Stylization
Flux	Photoreal, text	Newer / less ecosystem
Stable Diffusion	Customizable, batch	Setup complexity
Adobe Firefly	Commercial-safe (no copyright issue)	Mid quality

매 prompt structure (commercial)

[Subject]: 매 product / person.
[Setting]: 매 environment.
[Lighting]: 매 studio / natural / dramatic.
[Style]: 매 photorealistic / minimalist / editorial.
[Composition]: 매 rule of thirds / centered.
[Quality]: 매 4k, professional photography.
[Negative]: 매 blurry, watermark, deformed.

Midjourney V7 Draft Mode

매 4× faster, 매 1/4 cost.
매 multiple iteration의 affordable.
매 winner 의 high-quality re-render.
매 commercial pipeline 의 game-changer.

Typography (handle separately)

매 most AI 의 text 의 weak (DALL-E 3, Flux 의 better).
매 --no text (Midjourney).
매 후작업: Photoshop, Figma 의 text overlay.

Brand consistency

Brand Consistency Maintenance 참조.
매 sref / cref / IP-Adapter / LoRA.

License + copyright

Midjourney: 매 paid tier 만 commercial.
DALL-E 3: OpenAI ToS — 매 commercial OK.
Stable Diffusion: 매 model license 의 dependent.
Adobe Firefly: 매 commercial-safe (Adobe Stock train).
Getty / Adobe Stock: 매 indemnification.

매 안전 (SFW + IP)

매 prompt 의 celebrity / IP 의 X (legal).
매 generated 의 trademark check.
매 client 의 legal review.

💻 패턴

Product mockup prompt (e-commerce)

Professional product photography, 
[product description], 
on seamless white background, 
soft studio lighting from front-left, 
shot with 50mm lens at f/2.8, 
high-end commercial photography style, 
sharp focus on product, 
8k resolution, hero image
--ar 1:1 --stylize 100 --no text

Ad creative (lifestyle)

Lifestyle photography of [target demo] enjoying [product],
shot at [location], golden hour lighting,
authentic candid moment, shallow depth of field,
shot with 35mm lens, 
professional editorial photography style
--ar 16:9 --stylize 200

Banner (text-heavy via DALL-E 3)

Marketing banner for "ACME Cloud Storage",
clean modern design, blue and white palette,
include the text "Save 50% on Annual Plans" prominently,
geometric patterns, professional tech aesthetic,
16:9 aspect ratio

Brand consistency (Midjourney)

[product description]
--sref https://my-cdn/brand-style-1.jpg
--sref https://my-cdn/brand-style-2.jpg
--sw 200
--ar 4:5 --stylize 100

Draft → Final pipeline (Midjourney V7)

// 매 1. 매 draft 의 N variations (cheap)
const drafts = await mj.generate({
  prompt: brandPrompt,
  draft: true,
  count: 16,
});

// 매 2. 매 human 의 select
const selected = await humanSelect(drafts);

// 매 3. 매 high-quality re-render
const finals = await Promise.all(
  selected.map(d => mj.upscale(d, { quality: 'high' })),
);

Stable Diffusion batch (Flux + IP-Adapter)

from diffusers import FluxPipeline
import torch

pipe = FluxPipeline.from_pretrained('black-forest-labs/FLUX.1-dev', torch_dtype=torch.bfloat16).to('cuda')
pipe.load_ip_adapter('XLabs-AI/flux-ip-adapter')

brand_ref = Image.open('brand_style.jpg')
products = ['shoe', 'jacket', 'bag']

results = []
for product in products:
    img = pipe(
        prompt=f'professional product photo of {product}, white background, studio',
        ip_adapter_image=brand_ref,
        ip_adapter_scale=0.6,
        guidance_scale=3.5,
        num_inference_steps=30,
    ).images[0]
    results.append(img)

Adobe Firefly (commercial-safe)

import requests

# 매 Firefly API
response = requests.post(
    'https://firefly-api.adobe.io/v3/images/generate',
    headers={
        'Authorization': f'Bearer {token}',
        'X-API-Key': api_key,
    },
    json={
        'prompt': 'professional product photography of red sneakers',
        'numVariations': 4,
        'size': {'width': 2048, 'height': 2048},
        'styles': {'presets': ['photo']},
        'seeds': [42],
    },
)

Negative prompt (avoid commercial defects)

NEGATIVE_DEFECTS = "blurry, low quality, jpeg artifacts, watermark, signature, " + \
                   "deformed hands, extra fingers, distorted face, asymmetric eyes, " + \
                   "amateur, snapshot, low resolution"

License + provenance check

def commercial_use_check(image, generation_metadata):
    issues = []
    
    # 매 platform license
    if generation_metadata['platform'] == 'midjourney':
        if not generation_metadata['user_subscription'] == 'paid':
            issues.append('Midjourney requires paid tier for commercial.')
    
    # 매 model license
    if 'sd' in generation_metadata['model']:
        if 'creativeml' in generation_metadata.get('license', ''):
            issues.append('Check CreativeML license restrictions.')
    
    # 매 trademark check (manual)
    if contains_logo(image):
        issues.append('Manual trademark review required.')
    
    # 매 C2PA provenance
    if not has_c2pa_manifest(image):
        issues.append('Missing C2PA — consider adding for transparency.')
    
    return issues

Production batch script

import csv

def batch_commercial(brief_csv):
    with open(brief_csv) as f:
        briefs = list(csv.DictReader(f))
    
    for brief in briefs:
        prompt = build_prompt_from_brief(brief)
        drafts = generate_drafts(prompt, n=8)
        save_drafts_for_review(brief['id'], drafts)
        
    print(f'Generated {len(briefs)} draft sets. Review at /drafts.')

Cost monitoring

class GenerationCostTracker:
    def __init__(self):
        self.costs = []
    
    def log(self, platform, mode, count):
        unit_cost = COST_TABLE[platform][mode]  # 매 매 image
        self.costs.append({
            'platform': platform,
            'mode': mode,
            'count': count,
            'cost': count * unit_cost,
            'date': datetime.now(),
        })
    
    def daily(self):
        today = [c for c in self.costs if c['date'].date() == date.today()]
        return sum(c['cost'] for c in today)

🤔 결정 기준

사용처	Tool
Hero image	Midjourney + draft pipeline
Logo / text-in-image	DALL-E 3
Photoreal product	Flux
Bulk variation	SD + IP-Adapter
Commercial-safe	Adobe Firefly
Brand mascot	Midjourney `--cref` or LoRA
Concept art	Midjourney

기본값: Midjourney (draft → final) + DALL-E 3 (text) + Firefly (safe). 매 brand = sref / LoRA.

🔗 Graph

부모: AI 이미지 생성 (AI Image Generation) · Marketing
변형: Brand Consistency Maintenance · ChatGPT_Emoticon_Prompt_Engineering
응용: Midjourney · DALL-E · Flux · Adobe-Firefly · Stable-Diffusion
Adjacent: Authenticity · Arts · CFG 스케일(Classifier-Free Guidance Scale)

🤖 LLM 활용

언제: 매 brand asset 매 production. 매 ad campaign. 매 catalog. 매 social media. 언제 X: 매 fine art (artist 의 expression). 매 license unclear (legal review).

❌ 안티패턴

No license check: 매 ToS violation.
Celebrity / IP 의 prompt: 매 lawsuit risk.
Text-heavy on Midjourney: 매 후작업 의 X.
Single platform monoculture: 매 same-look.
No human review: 매 brand 의 off.
No draft mode (Midjourney V7): 매 cost 폭발.
No C2PA disclosure: 매 audience 의 trust.

🧪 검증 / 중복

Verified (Midjourney docs, OpenAI ToS, Adobe Firefly commercial-safe claim).
신뢰도 B.
Related: Brand Consistency Maintenance · Authenticity · CFG 스케일(Classifier-Free Guidance Scale) · Arts · AI-backed-Image-Generation-Workflow.

🕓 Changelog

날짜	변경
2026-04-30	Auto-mapped
2026-05-08	Phase 1
2026-05-10	Manual cleanup — platform comparison + draft mode + 매 product / batch / Firefly code

9.3 KiB Raw Blame History Unescape Escape