Key takeaway — AI video models don't understand "make it look cool," but they do understand "low-angle medium shot." To get the ad video you have in mind, you first need the vocabulary to direct it. The xbrush Academy organizes that vocabulary into 169 concepts across 13 areas. This post condenses the map into one page; the next five posts cover 20 practical tips and real, publicly shared prompt examples.
Why start with concepts
Most people hit the same wall the first time they generate video with AI. The scene in your head is clear, but every result comes out a little different, and you can't tell what to fix.
The usual cause is that the direction is written entirely in adjectives. Words like "sensual," "premium," or "warm" mean different things to different people, and even more so to a model. Production and editing terms, on the other hand, are words the model has seen countless times in training. Name the thing and the field of candidates narrows — and when the result is wrong, you have a handle to turn.
That is why the xbrush Academy treats the concept map as its vocabulary course, one of three courses covering ordering, vocabulary, and judgement. Before you can tell an agent what you want, you need a name for it.
The concept map at a glance — 13 areas
The 13 areas are easiest to read as a flow: plan → make → refine → ship → operate. The table below picks three or four concepts per area worth learning first, each summarized in one line. All 169 concepts are in the xbrush Academy Concept Map.
Area (concepts) | Learn these first | In one line |
|---|---|---|
1. Ad Planning (16) | Brief · Selling point/USP · Key message · CTA | Write down who it's for, what to say, in one line, and what you want them to do |
2. Ad Video Structure (16) | Hook · Ad length (6 · 15 · 30 sec) · Structure formulas (PAS) · End card | Grab attention in the first 2–3 seconds and fold the story to fit the length |
3. Film Grammar (15) | Shot size (ECU–WS) · Camera angle · Camera movement · Continuity | How to call the camera by name, and how to make shots connect |
4. AI Generation · Image (19) | Reference input · Consistency · Product fidelity · Genericness ("AI look") | The vocabulary for images that become first frames and references |
5. AI Generation · Video (19) | Image to video (I2V) · Motion prompt · Start/end frame · Temporal consistency | How to direct motion and reduce drift between shots |
6. Audio (15) | Voice-over (VO) · BGM · Sound effects · Ambience · Ducking | How to direct speech, music, and diegetic sound separately, then mix them |
7. Text · Graphics (12) | Copy hierarchy · Subtitles · Safe zone · Logo usage rules | Keeping on-screen text from being cropped or covered |
8. Editing · Post (9) | Timeline · Tracks/layers · Speed adjustment · Render/export | Assembling generated clips into one film |
9. Technical Specs (11) | Aspect ratio (1:1 · 4:5 · 9:16 · 16:9) · Resolution · Thumbnail/cover frame | Matching the specs a placement requires from the start |
10. Media · Platforms (8) | Placements (Feed · Stories · Reels · Shorts) · In-stream ads · Platform ad policies | Where it runs decides the length, ratio, and first frame |
11. Performance Measurement (7) | 3-second view rate/hook rate · CTR · Creative testing (A/B) | What to watch after posting, and how to fix the next one |
12. Collaboration · Delivery (12) | Draft for review · Sign-off/feedback · Revision rounds · Delivery package | The units of agreement when working with clients and teams |
13. Rights · Law · Ethics (10) | Copyright in AI-generated work · Likeness rights · AI content disclosure · Deceptive advertising | The lines to check before anything goes live |
You don't need to memorize all 169 up front. The concepts that show up in prompts most often come from areas 2, 3, 5, and 6 (structure, film grammar, video generation, audio), and most of the tips in this series use vocabulary from those four.
How the prompt changes when the vocabulary does
Put the same intent side by side — once in adjectives, once in concept names — and the difference is obvious.
Intent | Written in adjectives | Written in concept names |
|---|---|---|
Make the product stand out | Show the product from a stylish angle | Close-up, 15° low angle, locked off. Product label centered in frame |
Warm mood | Warm and bright, please | 4pm, side light from the window on the left |
Short and punchy | Make me a nice coffee ad | 15 seconds, 4 shots. Cut every 3–4 seconds; shot 1 is the hook |
For Reels | I'll crop it vertical later | 9:16 vertical. Leave the bottom 20% clear for captions |
The right-hand column isn't long. Instead, every word is a handle you can adjust. If the result feels too dramatic, change "15° low angle" to "eye level"; too cold, change "side light" to "backlight" — one word at a time.
Concepts → tips → practice: the Academy path
The three xbrush Academy courses each do a different job.
Concept Map — what to call things (vocabulary)
20 Tips — how to place the order (ordering)
10 Hands-On Steps — what to fix after you see the result (judgement)
The hands-on track starts with turning one product photo into an ad shot (L1), then moves through one idea into four placements (L3), A/B variants (L5), turning a photo into video (L6), a 15-second vertical ad with sound and text (L7), an ad video with Cinema (L8), and finally publishing and checking metrics 7 days later (L9). Anyone can read the lessons without signing in; making things requires a sign-in, and signing up gives you 100C in practice credits.
What this series covers
The next four posts walk through the Academy's "20 Tips for Directing That Lands" in its four groups. Every tip comes with why it works and a short excerpt from a real, publicly shared AI video prompt that applies it. The final post maps out where you can find prompts like these.
Before you order — length, purpose, aspect ratio, budget, references (tips 1–5)
The shape of a prompt — three structures, timecodes, one action per shot, block names (tips 6–10)
Locking what's on screen — people, camera, light, color, negatives (tips 11–15)
Sound, dialogue, and the ending — dialogue, lip sync, sound vs. music, the final frame, how to revise (tips 16–20)
Where to find great ad video prompts — official guides, open-source libraries, creator communities
Frequently Asked Questions
Do I need to know filmmaking terms to make AI ad videos?
You don't need all of them, but knowing shot size, camera angle, and camera movement makes results far more stable. AI video models have learned these terms, so they respond much more precisely to a name like "low-angle medium shot" than to an adjective like "cool."
What's in the xbrush Academy Concept Map?
It organizes 169 concepts into 13 areas: ad planning, ad video structure, film grammar, AI image and video generation, audio, text and graphics, editing, technical specs, media and platforms, performance measurement, collaboration, and rights and ethics. Each concept comes with a practice task you can try right away with the xbrush.ai agent.
Which areas should I study first?
Start with the four areas that show up in prompts most often: ad video structure, film grammar, AI video generation, and audio. Then look at technical specs and media to design videos that fit their placement from the start.
Is the Academy free?
Lessons, concepts, and tips can be read without signing in. Signing in is required once you start making images or videos in the hands-on lessons, and new accounts receive 100C in credits to practice with.
In what order should I read this series?
Skim the concept map in this post, then read tips 1–5 (before you order), 6–10 (prompt structure), 11–15 (locking the frame), and 16–20 (sound and endings) in order. If you want more real prompt examples, see the source guide in the final post.