AI Prompting Tips
Get Pro-Level Results Fast

1) Golden Principles of Prompting

• Specific beats generic — state the important details: goal, audience, constraints, style.
• Give examples — 1–2 sample outputs dramatically improve accuracy.
• Define the output format — enforce structure (markdown, JSON, table) for clean, usable results.
• Request step-by-step — use "break into steps", "verify", or "self-review".
• Iterate quickly — run 2–3 rounds with feedback; never expect perfection on the first try.

2) Prompt Anatomy (RTCC-FES)

Role • Task • Context • Constraints • Format • Examples • Steps

Role: You are a senior content editor for a tech brand.
Task: Write a 600-word article on AI benefits for small businesses.
Context: Audience is small business owners, friendly tone, avoid jargon.
Constraints: Max 600 words, use H2-H3 headings, include 3 bullets.
Format: Markdown.
Examples: Desired title — "AI for SMBs: Save Time, Grow Revenue".
Steps: 1) Outline  2) Draft  3) Quick self-review  4) Final version.

3) Universal Text Template

Role: <profession>
Goal: <what needs to be produced>
Audience: <who> | Tone: <friendly/formal/technical>
Constraints: <length, keywords, restrictions>
Format: <markdown/table/json>
Example: "<short sample of target output>"
Steps: <step-by-step>

Output only: <format>. Do not write explanations outside the format.

Example — SEO Article Outline

Role: SEO strategist.
Goal: Create an outline for "AI Prompting for Beginners".
Audience: beginner marketers. Tone: clear and warm.
Constraints: 1200–1500 words when written later.
Format: Markdown (H2/H3 + bullets).
Steps: 1) Research intent  2) Build headings  3) Add FAQ.

Output: outline only.

4) Image Template (Midjourney / Leonardo)

Structure: Subject • Setting • Composition • Lighting • Style • Lens • Technical Details

// Midjourney (v6+/Niji)
<main subject>, <action>, <setting>, cinematic composition,
lighting <rim/volumetric/softbox>, color grading teal-orange,
lens 50mm f1.8, ultra-detailed, photorealistic, DOF,
--ar 16:9 --v 6 --style raw

// Leonardo — Warkop AI brand style (neon blue)
A futuristic humanoid robot presenter at a news desk,
studio lighting, neon blue glow highlights, glossy black materials,
bokeh background, high detail skin/material, 8k upscale,
ratio 16:9, sharp focus

Quick Presets (copy–paste)

• cinematic, volumetric light, anamorphic bokeh
• hyperreal textures, micro-details, subsurface scattering
• global illumination, physically based rendering

5) Video Template (Veo + JSON Flow)

Use a narrative brief + shot flow (each shot ≈3–5 sec) + camera & audio instructions.

// Short Veo prompt (narrative)
Create a 10-second cinematic clip of a robot chef plating seafood fried rice.
Style: realistic, studio kitchen, neon-blue accent, soft key light.
Camera: slow dolly-in, 35mm, shallow DOF.
Sound: gentle kitchen sizzle + soft ambient.
End with hero plate close-up, steam visible.

// JSON Flow skeleton
{
  "meta": {"duration": 10, "ratio": "16:9", "mood": "cinematic"},
  "shots": [
    {
      "id": "s1",
      "desc": "Wide studio kitchen, robot chef at wok, flames burst",
      "camera": "dolly-in, 35mm, f/2.0",
      "audio": ["sizzle_soft", "ambience_low"],
      "notes": "neon blue edge light"
    },
    {
      "id": "s2",
      "desc": "Close-up plating seafood fried rice, steam visible",
      "camera": "macro, 85mm, rack focus to plate",
      "audio": ["sizzle_close"],
      "notes": "highlight texture, appetizing"
    }
  ]
}

6) Character & Style Consistency

For recurring characters, write a short "visual ID card" and paste it into every image/video prompt.

// Short character bio (re-paste in every prompt)
Character: Angora Cat — A hyperrealistic cinematic portrait
of a fluffy white Angora cat, sitting gracefully on a velvet chair.
The cat has long silky fur, sparkling blue eyes, and a regal expression.
Soft studio lighting, shallow depth of field, ultra detailed whiskers
and fur texture, 8k resolution, cinematic mood, photorealistic.

7) Debugging Prompts (Quick Fix)

1. Diagnose — what went wrong? (color, face, composition, format?)
2. Hypothesize — likely cause (too little detail, style conflict, wrong lens?)
3. Fix — add constraints & examples; change lens/angle; split into 2 stages

// Debug template
Problem: <brief summary>. Likely cause: <technical/content>.
Revised prompt: <new version>.
Check: 3 success criteria (<c1>, <c2>, <c3>).

8) Prompt Cheatcodes

• "Show steps & verify" — ask the model to check its output against a checklist.
• "Critique mode" — request 3 critique points first, then ask for the final version.
• "Style transfer" — "rewrite in the style of <writer/brand>, keep points X and Y".
• "Output guard" — "respond ONLY in valid JSON".
• "Batching" — "generate 5 short variants (max 120 characters each)".

9) Quick FAQ