PDF Course + Discord · July 2026 Edition

Prompt Engineering for AI Video: PSALM Framework

English prompts that produce a usable clip on the first attempt. No 50 retries, no guessing. The PSALM framework, a 48-page English prompt bank, camera control for 12 moves, and acronym phonetics for B2B, SaaS, and ROI. Works with Sora 2, Veo 3, Runway, Kling, and LTX.

PSALM Framework 48-page English prompt bank 10 industries, ready-made templates Discord 24/7 Lifetime access

Join for $59 → See the PSALM framework

One-time payment, no subscription · 14-day withdrawal right under consumer protection law

I know why your AI video prompts frustrate you

4 reasons your prompts are not producing what you imagine

You generate the same clip 30 times and it never quite lands

You type your idea, get a cinematic blur. You change one word, you get the same blur with a slightly different detail. After 30 attempts your daily credits are gone and you have one clip that's just barely publishable. You're burning a subscription on noise.

You have no system — every prompt is a fresh gamble

Every session starts from scratch. No structure means no repeatability: a prompt that worked last Tuesday produces something completely different today, and you can't explain why. Without a framework, you can't improve — you can only keep guessing and hoping.

You don't know how to describe camera movement, so you skip it

AI generates a static frame or an unwanted zoom that nobody ordered. No camera control means no dynamics, no storytelling, no difference between your clip and the 100 others using the same prompt. Clients ask 'can't it look like a TV commercial?' and you have no answer.

Lip-sync sounds robotic — acronyms come out wrong

You type 'the presenter says: our B2B SaaS platform delivers ROI' and the model pronounces every acronym letter-by-letter in an unnatural machine voice. Business-facing content falls apart without knowing which phonetic spellings force the model to pronounce things correctly.

Why this matters: Prompt engineering for AI video is not an art — it is an engineering discipline. Sora 2, Veo 3, Runway, and Kling all respond to structure, not to creativity alone. Without a system you burn credits, time, and money. The PSALM framework taught in this course has already given 200+ creators control over their output instead of a guessing game. See how PSALM works →

What you will learn about prompt engineering for AI video

7 concrete prompting skills you will leave the course with

1
PSALM Framework — the complete prompt structure
Plan, Setting, Actor, Light, Motion. Five categories that organise every prompt. Works for Sora 2, Veo 3, Runway, and Kling. Once you have PSALM, you write prompts in 3 minutes instead of 30, and the result lands on attempt 1-3.
2
48-page English prompt bank for 10 industries
Restaurant, beauty, e-commerce, real estate, fitness, education, B2B, legal, salon, hospitality. Five ready-made templates per industry — fill in your brand and send. Per-model variants (Sora / Veo / Runway / Kling) in a single reference table.
3
Acronym phonetics for B2B, SaaS, ROI, KPI, and 77 more
A table of 80 common business acronyms with their prompt-ready phonetic spellings. Your clip says "sass" instead of "S-A-A-S" and "ar-oh-eye" instead of letter-by-letter. Clients in B2B and SaaS markets hear the difference immediately.
4
Camera control — 12 moves (dolly, orbit, crane, zoom)
Dolly, orbit, crane, zoom, pan, tilt, tracking, push, pull, arc, jib, handheld. Each move described with per-model syntax. Your clips have the dynamics that nobody gets when they skip camera instruction entirely.
5
Describing emotions in English prompts
A 60-word emotion vocabulary in prompt-friendly form. Not "smiling person" but "a subtle upturn at the corners of the mouth, eyes slightly narrowed, face relaxed." Actors in clips look like humans, not emoji. Sora 2 and Veo 3 both respond to this level of specificity.
6
Prompt iteration workflow — when the first shot misses
5-step debug: identify exactly what is wrong, change one element at a time, log in the workbook, and after 3 failed iterations switch models. This workflow preserves your daily generation credits. Without it you repeat the same mistakes until your credits run out.
7
Prompt chaining — consistency across a clip series
Five clips in one project, same character, same location, same visual style. The technique for multi-shot ads where a client wants "the same presenter in 5 situations." Without prompt chaining each clip looks like a different film; with chaining you deliver a coherent series.

The PSALM framework, step by step

5 letters that turn your prompts from chaos into production

PSALM is an acronym for 5 prompt categories. The sequence matters: AI video models are trained on scene descriptions that follow a particular structure. When you write in that same order, the model understands you better, generates what you actually want, and not just what your words loosely resemble. Below is a quick overview — in the course each letter gets its own dedicated chapter.

Plan

What type of scene, what goal, what format (ad, explainer, vlog, documentary). Without a stated goal a prompt is just rambling; with one, every word earns its place. This is the first and most frequently skipped letter.

Setting

Location, time of day, weather, props. This is where you describe the world. A rooftop bar in downtown Manhattan, overcast midday light, neon signage reflected in rain-slicked concrete. Specifics, not "a nice place."

Actor

Who, what they look like, what emotion they carry, what they do, what they say. This is where acronym phonetics and emotion vocabulary matter most. Without a precise emotion description, actors look like clip-art rather than people.

Light

Light and colour treatment. Cinematic golden hour, soft daylight, neon, hard backlit. Lighting determines mood; mood determines whether the client approves the clip or sends it back with notes.

Motion

Camera motion and in-frame action. The 12 camera moves (dolly, orbit, crane, zoom), plus actor motion (walks toward camera, reaches for object, turns away). Without Motion your clip is a postcard; with Motion it is a film.

→

Output

Once you have PSALM, you write prompts in 3-5 minutes and get a production-ready result in 1-3 attempts instead of 30. With a system instead of inspiration, you produce ads faster than a traditional creative agency.

Course curriculum

6 modules, 168 pages — PSALM runs through all of them

The course is built around one idea: give you structure instead of inspiration. Each module extends the PSALM framework with a new layer — from your first prompt in 10 minutes, through acronym phonetics, camera control, and iteration, all the way to prompt chaining for clip series. In English, with English examples, for international markets.

You will not find chapters explaining what AI is or a history of Sora 2. The course assumes: you know AI video exists, you have tried it, and prompts are frustrating you. This course is a concrete recipe for prompts that deliver results. Every chapter ends with an exercise; every exercise has a model answer in the workbook; every answer has 3-5 per-model variants (Sora 2, Veo 3, Runway, Kling, LTX). No theory without practice; no practice without a ready-made template.

Module 1 — PSALM Framework: the Foundation of Every Prompt

28 pages

→ PSALM decoded: Plan, Setting, Actor, Light, Motion — why this sequence matters
→ Anatomy of a production prompt: from 15 words to 80, and when more hurts
→ First PSALM prompt in 10 minutes: generate a clip that lands on the first try

Module 2 — English Prompt Language: Precision and Register

32 pages

→ Why 'make it cinematic' fails — how to actually direct an AI video model with words
→ Acronym phonetics for English content: B2B, SaaS, ROI, KPI, AI — table of 80 business terms
→ Describing emotions in prompt-friendly language: 60-word vocabulary for natural performances
→ English lip-sync: sentence length, pauses, intonation cues that hold synchronisation

Module 3 — Camera Control and Composition

22 pages

→ 12 core camera moves (dolly, orbit, crane, zoom, pan, tilt, tracking, push, pull, arc, jib, handheld)
→ Frame composition in a prompt: rule of thirds, leading lines, depth layers
→ Shot types in a prompt: wide, medium, close-up, detail, macro — how to specify each
→ Camera move syntax per model: Sora 2, Veo 3, Runway, Kling each read moves differently

Module 4 — Prompt Iteration: Workflow for a Missed First Result

18 pages

→ 5-step prompt debug: pinpointing exactly what is wrong
→ One-change rule: why editing three things at once makes diagnosis impossible
→ Iteration workbook: log changes so you never repeat the same failed attempt after 50 tries
→ When to give up on a prompt and switch models: not every clip can be salvaged in Sora 2

Module 5 — English Prompt Bank for 10 Industries: Ready-Made Templates

48 pages

→ Restaurant, beauty, e-commerce, real estate, fitness, education, B2B, legal, salon, hospitality
→ 5 templates per industry, ready to fill in with your own content
→ Per-model variant (Sora 2 / Veo 3 / Runway / Kling) for every template
→ Negative prompts: what to avoid — list of 30 words that degrade generation quality

Module 6 — Advanced Prompts: Style, Consistency, and Chaining

20 pages

→ Visual style in a prompt: cinematic, documentary, commercial, vlog, retro, neon
→ Character consistency across clips: maintaining face, clothing, and environment
→ Prompt chaining: building a series of 5 clips that look like one continuous scene
→ Format-specific prompts: 9:16 TikTok, 1:1 Instagram, 16:9 YouTube

Start the Prompt Engineering Course now

Guessing vs the PSALM system

Your workflow before PSALM vs after PSALM

Data from 50 comparative tests: 25 prompts written intuitively vs 25 prompts written in the PSALM structure. Full methodology documented in Module 3, with raw data available in the workbook. Tests ran across 4 models (Sora 2, Veo 3, Runway Gen-4, Kling 3) to rule out single-tool luck. Results were consistent across every model — the PSALM structure outperforms guesswork regardless of which generator you use.

Parameter	Guessing	PSALM System
Attempts per usable clip	20-50 attempts	2-4 attempts
Time to write a prompt	15-30 min, constant rewrites	3-5 min with PSALM structure
Lip-sync quality	Robotic, acronyms mispronounced	Natural delivery, phonetic fixes
Camera control	Static frame or random zoom	12 moves on demand
Consistency across a clip series	Each clip looks like a different film	Prompt chaining, coherent series
Daily generation credits	Burned in 2h on failed experiments	Enough for 3-5 projects
Time from brief to finished clip	4-8 hours	30-60 minutes

PSALM is not magic — it is structure. It works because AI video models are trained on scene descriptions that follow specific categories (setting, actor, light, motion). Writing in that order means the model understands you better and delivers what you want in 1-3 attempts. See also our full AI video courses hub.

From students who mastered prompt engineering

What they say after learning prompt engineering with this course

"Prompt engineering was the turning point. Before, I was guessing. After the course I have the PSALM framework and I know exactly what to write. I went from 30 attempts per clip down to 3. My daily credits finally last the whole day."

OD Ola D. Marketing specialist, creative agency

"First client project two weeks after starting the course. I took a template from the prompt bank, filled in the client's brand details, and sent it over. The client couldn't believe it was AI-generated — the structured English prompts give a polished, professional look that generic guesswork never produces."

AK Anna K. Video freelancer

"As a designer I had no background in writing prompts. PSALM was like Photoshop layers — a structure that suddenly made sense. The B2B acronym phonetics table gave me an edge over agencies who were still getting mispronounced product names in their clips."

KM Kamil M. Former graphic designer, AI freelancer

Why prompt language and structure both matter

A good idea is not enough — structure is what the model actually reads

English is the strongest language for AI video prompting because all five major models (Sora 2, Veo 3, Runway, Kling, LTX) were primarily trained on English-language descriptions. Writing in English does not mean sacrificing specificity or cultural context — it means writing precise, structured English that the model can interpret without ambiguity.

The common mistake is treating a prompt like a creative brief written for a human: "a warm, cinematic scene in a coffee shop." Models do not interpret this the way a director would. They pattern-match against training data. "A warm, cinematic scene" is too abstract; the model fills in the gaps with the most frequent match in its training set — which is almost always American generic. Specifics override generics: "a narrow espresso bar in Rome at 7 AM, harsh morning light through a single east-facing window, worn marble counter" tells the model exactly what training examples to draw from.

The second layer is lip-sync. For B2B and SaaS content, acronym pronunciation is a visible quality signal. A presenter who says "ess-ay-ay-ess" instead of "sass" breaks immersion for any industry-familiar viewer. Phonetic notation in the prompt corrects this at the source — it is a 30-second fix that separates amateur output from professional delivery.

The third layer is camera intention. Clients who commission video ads have seen thousands of professional productions. They have a strong intuition for what "looks right" — and a static, undirected frame almost never does. Adding explicit camera motion (dolly, crane, orbit) and lighting intent to a prompt is the difference between a clip a client approves on first viewing and one that comes back with the note "can it be more... dynamic?"

About the author

Łukasz Kowalski, AI Video Course Creator

I was writing prompts for AI video before anyone was using the term "prompt engineering." The PSALM framework emerged from notes taken across 2,000 generated clips in Sora 2, Veo 3, Runway, and Kling. I catalogued recurring patterns, tested them with 200 course students, and the result is a system that reduces attempts per clip from 30 down to 3. More about the author →

PSALM workflow example

From brief to finished prompt in 5 steps

Let's say a client orders a 10-second ad for a specialty coffee shop in Brooklyn. Here is how PSALM converts that brief into a finished prompt. This is exactly the same workflow taught in Module 1, plus 15 other industry examples in the workbook.

Step 1 — Plan

What is this scene, what is the goal. "10-second ad for a specialty coffee shop in Brooklyn, vertical 9:16 for Instagram Reels, goal: drive foot traffic during morning rush hour, tone: warm, unhurried, artisanal." This is not yet a prompt — it is your own brief to yourself.

Step 2 — Setting

Describe the place. "Interior of a small specialty coffee bar in Williamsburg, Brooklyn, early morning 7 AM, exposed brick walls, reclaimed wood counter, La Marzocco espresso machine, shelves of single-origin bags, warm Edison bulb lighting, condensation on the front window, street just visible through the glass." Concrete details — no generic "coffee shop."

Step 3 — Actor

Who and what they do. "Barista, early 30s, dark hair pulled back, white apron over a chambray shirt, focused concentration while pulling a double espresso, slight smile as the crema surface forms, eyes briefly flicker to camera — a moment of shared ritual." Specific emotion, not "happy barista."

Step 4 — Light

Light and colour. "Warm golden morning light through the front window backlighting the steam from the espresso, soft fill from overhead Edison bulbs, deep amber and brown palette, gentle lens flare, slight film grain, cinematic 2.35:1 crop bars even in vertical." A specific look, not "nice lighting."

Step 5 — Motion

Camera and in-frame motion. "Slow dolly forward over 8 seconds from a medium shot of the espresso machine to a close-up of the cup as the crema settles, then a gentle tilt up to the barista's face in the final 2 seconds, motion silky smooth, 24fps cinematic." Concrete camera direction instead of a static frame.

Result: A 250-word PSALM prompt, ready to paste into Sora 2, Veo 3, or Runway. The first generation hits in roughly 80% of cases; cosmetic refinements close in the next 1-2 iterations. The client gets a finished clip in 30 minutes instead of 6 hours of guesswork. That is the difference between billing $500 and billing $5,000 for the same deliverable. Start using PSALM →

Pricing

One payment. Lifetime access. The complete prompt engineering course.

JUL 2026 EDITION

Complete course

$59 $99

One-time payment, no subscription, no hidden costs

✓168-page PDF with the PSALM framework and full prompting workflow
✓12-page workbook with prompt iteration exercises
✓48-page English prompt bank for 10 industries (copy-paste ready)
✓Acronym phonetics table — 80 business terms (B2B, SaaS, ROI, KPI, and more)
✓Discord 24/7 — community + updates when models add new features
✓Lifetime access, 14-day withdrawal right under consumer law

Get the Prompt Engineering Course

Stripe · Card · Apple Pay · Google Pay. Access delivered in 1–2 minutes after payment.

FAQ

Common questions about the AI video prompt engineering course

Which AI video tools do the prompts in this course work with?

Sora 2 (OpenAI), Veo 3 (Google), Runway Gen-4, Kling 3, and LTX. The PSALM framework is model-agnostic — it teaches you prompt structure that you then adapt to each tool. The English prompt bank includes ready-made versions for all five tools, so you don't have to guess what each model understands differently.

Should I write prompts in English or in another language?

For most content — cinematic shots, commercial ads, explainers, product demos — English produces the best results across all five models. The course focuses on English prompting as the primary language, with a dedicated chapter on when and how to add localisation cues (specific cities, cultural details, regional signifiers) to make scenes feel authentic to a particular market.

How long does it take to learn prompt engineering for AI video?

You can master the PSALM framework in about 2 hours, and write your first production-quality prompt the same day. Full fluency — writing without consulting the template bank — takes 7 to 14 days of regular practice. The course provides 48 pages of ready-made templates, so even on day one you are generating clips that look like finished production.

Will these prompts work for both Sora 2 and Veo 3?

Yes — plus Runway, Kling, and LTX. Each of the five models has its own quirks. Sora 2 responds best to detailed camera descriptions; Veo 3 reacts well to emotion language; Runway needs motion blur descriptions; Kling works better with concise instructions. The English prompt bank has a column for each model for every template.

What is acronym phonetics and when does it matter?

When prompts include acronyms like B2B, SaaS, or ROI in a lip-sync clip, some models mis-pronounce them. Phonetic notation — writing 'bee-to-bee' or 'sass' in the prompt — corrects this. The course includes a table of 80 common business acronyms with their prompt-ready phonetic spellings, particularly useful for B2B and SaaS video content.

How do I specify camera moves like dolly, orbit, crane, or zoom?

Every model understands camera move names, but each responds slightly differently. Sora 2 works best with 'slow dolly forward'; Veo 3 responds to 'orbit shot 180 degrees clockwise'. The course covers 12 core camera moves with per-model syntax, plus example prompts showing how to combine camera motion with actor action inside the frame.

What do I do when my first result misses the mark?

The prompt iteration workflow is a 5-step process: identify exactly what is wrong (composition, motion, colour, emotion, camera), change one element at a time, log changes in the workbook, conserve daily generation limits, and after three failed iterations switch models. The entire Module 4 is dedicated to this methodology — without it you burn through daily limits repeating the same mistakes.

Do I need any technical background to write prompts?

No. PSALM is a five-category framework (Plan, Setting, Actor, Light, Motion) — no technical knowledge required. The only prerequisite is knowing what you want to show. The 48-page prompt bank gives you copy-paste templates where you substitute your own content, so you can start generating immediately.

Do I get access right after purchase?

Yes. After payment (Stripe — card, Apple Pay, Google Pay) an email with three PDF files and a Discord invite arrives within 1–2 minutes. The 48-page prompt bank is delivered as a separate copyable PDF (not a scan), so you can paste prompts directly into Sora or Veo without retyping.

July 2026 edition — current pricing ends at month close

Stop guessing prompts. Start writing them with the PSALM system.

You get a framework nobody else has written down this completely in English, plus a 48-page prompt bank ready to copy. From 30 attempts per clip down to 3. 14-day withdrawal right under consumer law — zero risk.

Get the Prompt Engineering Course now

One-time payment · Lifetime access · Delivered in 1–2 minutes

Course