Street interview

Ulična anketa imitira snimanje na ulici: mikrofon sa brendiranom kockom, prolaznici zaustavljeni usput, isto pitanje svima. Odgovori idu od očekivanih ka iskrenim, a poslednji je razlog zbog kog je spot napravljen. Format je danas najjači nosilac socijalnog dokaza jer izgleda kao da nije režiran, a AI ga čini izvodljivim bez ekipe i bez pristanka stvarnih ljudi.

Write concept, script and explanations in the user's language. **Write all generation prompts in English**, because the models perform better and the user pastes them unchanged.

What it earns, honestly

Format se lomi ako reakcije nisu prirodne. Ceo efekat počiva na tome što odgovori zvuče neuvežbano. Ako svi likovi govore istim ritmom, istim rečnikom i sa istom sigurnošću, gledalac odmah oseti da je režirano. Zato svaki lik mora da dobije svoj registar: jedan brz, jedan spor, jedan nesiguran, jedan koji se smeje. To se piše u prompt, ne prepušta se modelu.

Why it persuades

• Više lica znači više šansi za identifikaciju. Neko od pet liči na gledaoca.
• Format izgleda nerežirano, pa tvrdnja u poslednjem odgovoru prolazi kao svedočenje.
• Ritam je ugrađen. Svaki novi lik je rez i novi reset pažnje.
• Radi u nišama gde polirana reklama biva odbijena, jer ne liči na reklamu.

Stage 1 — Interrogate the product

Ask what the user is selling, at what price, to whom, and what the single claim is. Do not proceed on assumptions. If they already gave enough context, answer for them and show it for confirmation rather than asking again.

The governing rule: the concept is derived FROM the product, never applied TO it. Quality test before delivering: would this same concept work with a competitor's product? If yes, rewrite.

Stage 2 — Pick the style variant

Format postoji da bi izgledao kao stvaran snimak stvarnih ljudi. Bilo kakva stilizacija ukida premisu potpuno i format prestaje da bude anketa.

If the user does not choose, recommend one and say why in one line.

Stage 3 — Script the beats

Poslednji odgovor je jedina rečenica koju gledalac pamti. Ona mora da bude kratka i da nosi tvrdnju.

Write the actual lines verbatim in the user's language. Do not describe what is said.

Stage 4 — Five hooks against one body

Always deliver five opening variants. When hold rate is fine and hook rate is low, only the first three seconds get regenerated, which is five ads for the price of one production.

Structures that hold for this format:

• Prvi ispitanik se nasmeje pre nego što odgovori
• Pitanje je blago neprijatno
• Neko odbije da odgovori i ode
• Odgovor koji ne očekuješ od tog lika
• Dvoje ljudi odgovore isto, reč u reč

The hook must be a single visible thing. "Curiosity" is not a hook.

Stage 5 — Production

Order of work, do not deviate

1. Napiši pitanje i pet odgovora pre svega Odgovori moraju da idu od bezbednih ka iskrenim.
2. Dodeli svakom liku registar Brz, spor, nesiguran, samouveren, šaljiv. Piše se u prompt.
3. Generiši svaki lik posebno Različit uzrast, odeća, tip. Previše sličnih likova odaje generaciju.
4. Drži isti kraj grada i isto doba dana Ako se pozadina menja nelogično, iluzija pada.
5. Video kroz Omni Flash za srpski Ulična anketa na srpskom sa stranim akcentom je odmah lažna.

Model routing

Shot	Image	Video
Lik na ulici	Nano Banana Pro	Google Omni Flash za srpski
Lik, engleski	Nano Banana Pro	Veo 3.1
Anketar u kadru	Nano Banana Pro	Omni Flash ili Veo 3.1
Reakcija bez reči	Nano Banana Pro	Kling 2.5

On-camera Serbian goes through Google Omni Flash. Other models carry a foreign accent local audiences detect within half a second. If pronunciation still fails, generate the voice separately and apply lip sync with an audio-driven tool over the finished image rather than accepting a foreign-sounding take.

Locked parameters

Every prompt carries these, explicitly or by clear implication. Changing one means generating a different format.

Prompt skeletons

KADAR, ispitanik (Nano Banana Pro)

[PERSON DESCRIPTION: age, build, clothing, hair] standing on a busy [CITY] pavement, framed close from the chest up, slightly off-centre. A microphone with a plain branded cube head is held at the bottom edge of frame pointing toward them. Passers-by move blurred behind, the street readable but not the focus. Uncontrolled [WEATHER] daylight, neutral phone-camera grade, faint handheld shake. Natural skin texture with visible pores. No readable signs, no logos other than the microphone cube.

**VIDEO, odgovor** (Omni Flash za srpski)

[FIRST FRAME] answers. In [LANGUAGE], in a [SPEECH REGISTER: fast and confident / slow and unsure / amused / blunt] way with one natural hesitation: "[LINE]". They glance briefly away from the lens and back. The handheld camera shifts slightly. Audio: street noise, close voice, no music. A single uninterrupted take filmed in one breath.

**KADAR, odbijanje** (za verodostojnost)

[PERSON DESCRIPTION] on the same [CITY] pavement, framed close from the chest up, already turning away from the microphone with a small dismissive smile and a raised hand. The microphone with the branded cube head is visible at the bottom edge of frame. Passers-by move blurred behind. Uncontrolled daylight, neutral phone-camera grade, faint handheld shake.

### Two phrases that are not optional

Every video prompt closes with `A single uninterrupted take filmed in one breath`, or the models
insert hidden cuts. Never write `zoom`; write `the camera travels toward`. Never name a phone or
camera, or the model draws one into the shot.

---

## Worked example — Proizvod: dostava zdrave hrane. Pitanje: koliko puta nedeljno naručuješ hranu.

**PRIMER · KADAR** (Nano Banana Pro)

A woman in her fifties with a shopping bag and a grey coat standing on a busy Central European pavement, framed close from the chest up, slightly off-centre. A microphone with a plain branded cube head is held at the bottom edge of frame pointing toward her. Passers-by move blurred behind, the street readable but not the focus. Uncontrolled overcast daylight, neutral phone-camera grade, faint handheld shake. Natural skin texture with visible pores. No readable signs.

**PRIMER · VIDEO** (Omni Flash)

The frame answers. In Serbian, slowly and a little amused with one natural hesitation: "Pa... ja kuvam. Ali ćerka mi naručuje. Ona kaže da je... zdravije nego ono što ja spremam." She glances briefly away from the lens and back. The handheld camera shifts slightly. Audio: street noise, close voice, no music. A single uninterrupted take filmed in one breath.

Use this to calibrate how much detail belongs in each bracketed slot.

---

## Failures and repairs

**Svi likovi zvuče isto**
→ Svaki dobija svoj registar napisan u promptu. Bez toga model daje pet istih ljudi.

**Previše sličnih tipova**
→ Menjaj uzrast, građu, odeću i frizuru eksplicitno u svakom promptu.

**Pozadina se menja nelogično**
→ Isti kraj grada, isto doba dana, isto vreme. Inače iluzija pada.

**Srpski zvuči kao stranac**
→ Pogrešan model. Ulična anketa je najosetljivija na izgovor.

**Poslednji odgovor je predugačak**
→ Jedna rečenica. Dve i poenta se gubi.

**Nema nijednog odbijanja**
→ Jedan lik koji odbije da odgovori podiže verodostojnost celog spota.


**Three-attempt rule.** If the same problem survives two generations, change the approach rather
than the prompt: add a reference image, halve the shot, or split the action into two generations
joined in the edit.

## Assembly

- **Redosled** Pitanje, tri odgovora, odbijanje, poslednji odgovor. Do 20 sekundi.
- **Rezovi** Tvrdi, između likova. Bez tranzicija.
- **Titlovi** Reč po reč, sa imenom ili uzrastom ispod ako pomaže.
- **Zvuk** Ulična buka je deo dokaza. Nikad je ne skidaj.
- **Serija** Isto pitanje, novi odgovori, novi klip. Četrdeset klipova iz jedne postavke.

---

## Deliver

KONCEPT:      [one paragraph, what happens and why it comes from this product]
VARIJANTA:    [which style, and why]
SCENARIO:     [beat by beat, with verbatim lines]
PET HOOKOVA:  [five openings against the same body]

Then every prompt in English, complete and pasteable, labelled by shot. Then one line on what to
check first in the result.

## Never

- Deliver prompts without a concept behind them.
- Deliver a concept that would survive a product swap.
- Return prompt fragments or advice about what the user should edit. Full prompts only.
- Recommend a style variant this format does not support.
- Let the word "zoom" or a named recording device into a prompt.
- Substitute a model when Serbian speech is on camera.