← Community library
TV ad guide
How a short ad works as an ad: the opening, brand at both ends, voice-over, music, pacing, story shapes and the payoff.
Guide by Takeloom
Sign up to use it# TV Ad Craft Guide — 15-Second Widescreen Commercials
This document is injected context for the LLM when writing video ad scripts. It is the
companion to the realism guide: the **realism guide governs how each frame looks**
(camera, light, color, production quality); **this guide governs how the video works as
an ad** (opening, pacing, voiceover, music, story, payoff). Both apply at the same time —
they do not conflict. The target is a broadcast-quality 15-second TV spot in 16:9
widescreen (1280x720): the kind of commercial Nike, Apple, Canon, or Volvo would air
before a football match or run as a pre-roll — professionally produced, tightly
constructed, emotionally clear.
**Polished ≠ low stimulus.** A common failure mode is a "premium" video that is just
slow ambient B-roll: one lingering shot, room-tone audio, nothing said, nothing claimed.
That is not an ad — a viewer would not remember the brand five minutes later. Production
polish constrains the *rendering*, never the *energy or clarity* of the message.
**Other lengths and formats.** The numbers in this guide are for a 15-second 16:9 spot. For other
lengths, scale them: about one scene per 2.5–5 seconds, about 1.3–2.3 spoken words per second, a
music bed from 0s to the end, the payoff in the final 3 seconds. For a vertical 9:16 ad, the
vertical video guide takes precedence on framing, hook timing and text safe zones.
---
## The 3 Non-Negotiables
Every script must have all three. A script missing any one of them is not done.
1. **An engaging opening in the first 2–3 seconds.** TV viewers don't scroll away, but
they do tune out. Open on something with intrinsic interest: a striking image, a
moment of tension or motion, a bold spoken line, a question, a scene already in
progress. Never open with dead air — an empty establishing shot, a slow fade-in, or
someone silently doing a mundane task with no audio.
2. **A continuous audio spine.** Music bed and/or voiceover running through the whole
spot. Diegetic sounds (footsteps, shutter clicks, a door) are *texture on top* of the
spine, never the whole soundtrack. A 15s spot with only ambient sound is a rejected
script.
3. **A payoff in the final 3 seconds.** The core message lands: the strongest product
moment, the closing VO line, the brand. TV ads close with a clear brand moment — the
product hero shot and/or brand endcard belongs here. The viewer should be able to say
in one sentence what the ad claimed and who it was for.
---
## Brand Bookend — Establish the Brand Early AND Close on It
Strong reference TV ads show the brand at **both ends**: the product with its
logo visible in the opening, and the logo/product again at the close. This is a deliberate
convention, not an accident — match it.
- **Open branded.** Within the first 2–3 seconds the viewer must know whose ad this is.
The cleanest, most TV-authentic way is diegetic: the product already in use with its
brand mark legible — the logo on the device, the pack in hand, the wordmark on the
wrapper. This is *not* the "logo open" the Opening Patterns section warns against
(that means a static logo card floating on black); a branded product actually doing
something is a strong, correct open.
- **Close branded.** The final scene shows the product again as a real in-world shot with
the **brand logo prominently visible** (large in frame, label/logo toward camera, on the
product or packaging) as the endcard/tagline lands. The logo at the close is the single
most important brand moment — do not let it be small, turned away, or absent, and never
replace the real shot with a blank/black endcard card.
- **Same product both ends.** The product shown at the open, the hero, and the close must
be the same one as the reference image — same model and form factor — so the brand reads
consistently and the viewer never sees a different-looking unit.
---
## Opening Patterns (first 0–3s)
Pick one and commit. The opening earns the next 12 seconds.
- **Bold claim VO over the strongest visual**: "Some moments give you half a second."
Start mid-action, not at the beginning of the action.
- **Question or tension open**: a line or image that sets up a small problem the spot
will resolve ("How do you catch what never stands still?").
- **Human moment already in progress**: a scene the viewer instantly recognizes —
a birthday cake arriving, a sprint for the bus, a child mid-leap — joined mid-beat.
- **Striking product-adjacent image**: the world the product lives in, shot beautifully,
with the VO or music immediately signaling this is going somewhere.
Never open with: a slow reveal, an empty room, a static logo card on black, hands slowly
reaching for the product, someone waking up. These waste the seconds that decide whether
anyone listens. (Opening on the *branded product in real use* — its logo legible on the
device, pack, or wrapper — is not a "logo open"; that is the Brand Bookend above and is
required, not discouraged.)
---
## Voiceover and Speech
TV spots are carried by a written VO. Write it for the whole video unless there is a
deliberate creative reason for silence (rare).
- **Write every word verbatim** in the AUDIO lines. Never "VO about capturing moments".
- **Voice character — cast it to the audience, don't default.** The voice is a casting
decision driven by who the ad is for, the product, and the tone — not a house narrator.
Specify gender, a specific age band, timbre/texture, energy/pace, and emotional colour,
e.g. "female VO, late 20s, bright and playful, buoyant, close and friendly", "male VO,
30s, husky, grounded, unhurried", "female VO, 40s, cool and understated, dry wit", "male
VO, early 20s, energetic, fast, sunny". **Do not reach for the same "warm male, 40s,
measured" narrator every time** — two ads for two different audiences should have audibly
different voices. TV VO is composed and articulate, not rushed creator-speak, but its
gender/age/energy should visibly fit the target group (a playful summer-crowd spot and a
rugged trail spot are not read by the same person).
- **Density**: in 15 seconds a TV read fits roughly 20–35 spoken words. Fewer,
better-chosen words beat a dense read — the pictures and music carry meaning too.
- **Arc**: opening line that frames the idea → one development beat → closing line that
carries the core message. The last spoken line is the one people remember; the brand
slogan may be spoken there.
- Polished but human: plain words, active voice, no ad-speak clichés
("crafted for excellence" ❌, "the moment only happens once" ✅).
---
## Music
Always include a music bed from 0s to 15s.
- Describe it like a brief to a composer: genre, BPM feel, instrumentation, energy
curve — e.g. "warm piano and soft strings, ~90 BPM, builds gently through the middle,
opens up at the hero shot, resolves to a single held note under the endcard."
- The music should *move with the edit*: a build, a swell, or an instrument change at
scene transitions makes cuts feel intentional; the resolution lands with the payoff.
- Keep it plausible for the brand: premium brands use restrained, modern, emotive
beds — not EDM drops, not corporate ukulele, not stock "uplifting whistle".
---
## Pacing and Structure
- **15 seconds = 3–5 scenes.** Scene lengths of 2.5–5 seconds. TV pacing breathes more
than social media — shots are composed and held a beat longer — but three static
5-second shots is a slideshow, not a commercial.
- Vary the shots between scenes: location, distance, or framing must change noticeably
at every cut. Two adjacent scenes that could be stills from the same shot = cut one.
- **Write the cuts explicitly.** The video is generated in a single pass, so the edit
only exists if the prompt forces it: every scene after the first begins with the
literal words "Hard cut." and a completely new camera setup (different location or
position, AND different distance/height/side). Never write the camera or subject
moving *between* scenes — each scene is its own shot that starts cold, as if filmed
separately and cut together in the edit suite.
- **Within a scene**, one deliberate camera move is allowed (a slow push, a smooth pan
with the action) — this is TV, not handheld UGC. One move per scene, motivated by the
action, never stacked.
- **Something must happen in every scene** — an action completes, a line lands, the
product does its thing. No scene exists just to be looked at.
- Compress time confidently: morning → afternoon → golden hour in 15s reads perfectly.
Viewers fill gaps; you don't need transitions.
- The product's hero moment (its key benefit made physically visible) gets its own
scene, timed to the music/VO peak — usually scene 3 or 4.
- The final scene doubles as the brand payoff: strongest product visual, closing VO
line, music resolving. (The brand endcard text itself is added separately — do not
write on-screen text into the script.)
---
## Story Shapes That Work in 15s
Pick one shape; don't improvise structure.
1. **Slice of life**: 3–4 recognizable moments from one person's day or one occasion,
the product woven naturally through them, VO ties them together, payoff on the most
emotional beat.
2. **Problem → product → after**: 3–5s of the relatable pain, product enters, life
visibly better, closing line reframes it.
3. **Product demo played straight**: the benefit shown clearly and beautifully — the
thing the product does, made visible and undeniable, VO narrating what matters.
4. **Repetition with escalation**: the same action in 3 escalating contexts, one VO
line reframing each, the product constant throughout.
5. **Emotional montage**: short human moments unified by one idea and one music cue,
the product as the enabler, VO minimal and precise.
---
## Sound Design (on top of the spine)
Diegetic sounds add texture and punch — use them *with* music/VO, synced to the edit.
Good: a camera shutter on the beat, a door, a match strike, a ball hitting a glove.
Each sound is a physical event, short and clean. Avoid genuinely unpleasant sounds
(chewing, scraping, alarms) — but do not strip a scene to silence out of caution;
texture beats emptiness.
---
## Payoff / Ending
- The final scene is the strongest, not the leftovers: best product visual + the VO
closing line + music resolving.
- TV convention: the spot ends on a clear brand moment. The core message should land
here — implied by the story and nailed by the last line. Unlike social ads, speaking
the slogan verbatim at the end is normal and expected on TV, if delivered with craft.
- End on a completed beat — a held final image works on TV — never a fade-out of
someone ambling away mid-story.
- **The final shot is a real, fully-lit, in-world image — never a blank card.** The
brand endcard TEXT is added later as a transparent overlay *on top of* whatever the
final scene renders, so that scene must be an actual product shot in the ad's own
environment (the product with its logo clearly visible, in the world the spot lived
in). Do NOT end on a fade to black, a cut to a black/white/solid-color card, a studio
void, or an empty background waiting for text — a "black screen with only text" is a
broken ending. Describe the real closing image explicitly, and place the brand logo
prominently within it (large, label toward camera, on the product/packaging), so the
overlaid tagline lands over a genuine branded frame.
---
## On-Screen Text — how TV ads use it (written separately, not in the script)
Text overlays are designed in a separate pipeline step; the script itself contains no
on-screen text. But write the script knowing how TV text works, so the beats support it:
- TV ads do NOT subtitle or transcribe the voiceover. Text appears sparingly, to
**highlight a specific point**: a short claim or benefit ("Locks focus in 0.03s"),
a product name, and the closing tagline on the endcard.
- Typically 1–3 text moments in 15 seconds: at most one mid-spot highlight synced to
the hero moment, and a brand/tagline endcard over the final scene.
- This means the hero scene and the final scene should be composed with visual room
for text (a clean area of frame, a settled moment).
---
## Self-Check Before Submitting a Script
- Do the first 3 seconds give a viewer a reason to look up from their phone? What exactly?
- Read only the AUDIO lines: do they work as a piece of writing with a beginning,
middle and a memorable close? Is there music described from 0s to the end?
- Count the scenes: 3–5? Does something *happen* in each?
- Is the product's benefit physically visible in at least one scene (not just the
product being present)?
- Does the spot end on a clear brand payoff a TV viewer would recognize as the ending?
- Could this air on TV next to Nike/Apple-level commercials without embarrassment?