C.W.K.
Stream
Lesson 10 of 10 · published

Seeds, Rerolls, and the Practical Prompting Mindset

~18 min · prompting, control, l10

Level 0Spark
0 XP0/100 lessons0/14 achievements
0/200 XP to next level200 XP to go0% complete

피파 한 줄 정리: Practitioner mindset 한 줄: **direct, then curate**. Hero prompt 한 방이 아니라 batch generate → 선별 → refine → edit 루프. 사진가가 한 컷 찍고 안 가는 거랑 같은 이유.

Let's wrap up the prompting track with the mindset that separates people who get frustrated with AI image generation from people who consistently produce great work. It comes down to one principle: direct the model, then curate — don't command perfection in one shot.

The Photographer Mindset

Professional photographers don't take one photo and go home. They take hundreds — different angles, different moments, different exposures — and then select the best. The shoot (generation) and the selection (curation) are equally important parts of the process.

Generative AI works the same way:

The Practical Workflow:

1. DRAFT PROMPT          "Rough direction — subject, style, lighting"
       ↓
2. BATCH GENERATE        Generate 4-8 images with different seeds
       ↓
3. EVALUATE              Which ones have the best composition? Best mood?
       ↓
4. REFINE PROMPT         Adjust based on what you learned from the batch
       ↓
5. BATCH AGAIN           Generate 4-8 more with refined prompt
       ↓
6. SELECT + FIX          Pick the best, inpaint any problem areas
       ↓
7. DONE                  🖼️ Final image

Seed Strategies for Practical Work

Exploration phase: Use random seeds. You're fishing for compositions and interpretations. Don't commit to any single result yet.

Refinement phase: Lock a good seed, iterate on the prompt. You found a composition you like — now fine-tune the details without changing the underlying structure.

Variation phase: Lock the prompt, generate many seeds. Your prompt is dialed in — now find the perfect execution of it.

The Mindset Shifts

❌ Beginner Mindset

"I need to find the perfect prompt that gives me exactly what I want on the first try."

✅ Practitioner Mindset

"I'll write a good-enough prompt, generate a batch, learn from the results, and iterate. The final image will come from selection and refinement, not from one magical prompt."

❌ Expecting Obedience

"The model should do exactly what I tell it."

✅ Expecting Collaboration

"The model offers interpretations. I select the best ones and guide the next iteration."

❌ One-Shot Perfection

"If the first result isn't good, my prompt is wrong."

✅ Iterative Process

"The first batch gives me information. The second batch gets me closer. The third batch gives me the winner."

When to Stop Prompting and Start Editing

Here's the practical rule: if you've tried 3-4 prompt variations and 20+ seeds and you're still not getting what you want, the problem is probably not solvable by prompting alone. At that point:

  • Inpaint specific problem areas instead of regenerating everything
  • Use reference images (image-to-image) for stronger composition control
  • Try a different model that might handle your concept better
  • Break the scene into parts and composite them
  • Accept the "good enough" base and finish in Photoshop/post-processing
Key Takeaways
  • Direct, don't command. Generate batches, curate the best, refine through iteration.
  • Use exploration (random seeds) → refinement (locked seed, varied prompt) → selection (locked prompt, many seeds).
  • Curation is half the skill — developing your eye matters as much as prompt writing.
  • Know when to stop prompting and start editing — prompting alone can't solve every problem.
  • The "perfect image" usually comes from a workflow (prompt + generate + select + edit), not a single perfect prompt.

Code

예시 코드·python
# Exploration: cast a wide net
prompt = "abandoned greenhouse overtaken by wildflowers, morning light"
images = [generate(prompt, seed=random()) for _ in range(8)]
# Review all 8, identify which compositions and moods you like
예시 코드·python
# Refinement: lock the seed, vary the prompt
good_seed = 4829
prompts = [
    "abandoned greenhouse overtaken by wildflowers, morning light, 35mm film",
    "abandoned greenhouse overtaken by wildflowers, soft foggy morning, medium format",
    "abandoned greenhouse overtaken by wildflowers, golden hour, warm tones, shallow DOF",
]
images = [generate(p, seed=good_seed) for p in prompts]
# Compare how different prompt details affect the same base composition
예시 코드·python
# Final selection: perfect prompt, many seeds
final_prompt = "abandoned greenhouse overtaken by wildflowers, soft foggy morning, golden light filtering through broken glass, medium format film, warm and melancholic"
images = [generate(final_prompt, seed=s) for s in range(100, 120)]
# Pick the best from 20 candidates — one of them will be exceptional

External links

Exercise

관심 있는 prompt 골라 full practitioner 워크플로우 실행: random seed 8 (fast mode) → best seed 고정 + prompt 4 변형 → best prompt 고정 + 8 seed (full quality). 평소 워크플로우와 비교.

Progress

Progress is local-only — sign in to sync across devices.
이 페이지에서 버그를 발견하셨거나 피드백이 있으세요?문제 신고

댓글 0

🔔 답글 알림 (로그인 필요)
로그인댓글을 남기려면 로그인해 주세요.

아직 댓글이 없어요. 첫 댓글을 남겨보세요.