Parqlet, a guest parking platform for high-rise buildings, needed a LinkedIn ad for property managers and had no footage. Leo built it from the script, reference images and app screens: stock and generated scenes, an animated product, a licensed AI voice and caption files. The final 4K cut arrived in ten working days.
Parqlet is a small startup with a specific buyer. Its guest parking platform lets residents of high-rise buildings share the parking spots they aren’t using, and gives building management a real-time view of guest parking. The people who decide to bring it into a building are property managers and HOA boards, and the place Parqlet wanted to reach them was LinkedIn.
The brief: one landscape ad that explains guest parking to a property manager in under a minute, with nobody on camera.
What Parqlet sent
Margo Prais, Parqlet’s founder, shared a folder and a Figma file on day one: the script scene by scene, reference images for each shot, and the app screens. There was no footage to cut. That shaped the whole job, because every scene had to be built from stock footage, product animation or generated video, then held together by a voiceover.
We agreed the scope on the first day too. One landscape video for LinkedIn, ten working days, with the option to add vertical and square versions later. That kept the conversation on the story rather than on formats.
Finding the voice first
The voice went first, before any scene was cut, because it sets the pace for everything else. The first sample read the opening two lines. Margo’s notes were precise: it needed more energy, no long pauses, and no lift in pitch at the end of every sentence. The second voice kept the speed and dropped the pauses, and she approved it within a day.
While the voice was being chosen, the script got sharper. The opening now reads: “Your residents love having guests over. So every weekend, your front desk hears the same question: where can my guest park?” Two short lines put a property manager at their own front desk before the product appears.
That is the same order we use on every startup launch video: story and voice agreed first, so the cut has something to follow.
Stock where it fits, generated where it doesn’t
Each scene had a clear reference, and the first question was always whether real footage existed. Stock video won where it matched, because AI-generated people still tend to look generated, and a property manager notices. Where no stock clip showed the exact moment, such as a resident in a lobby asking the concierge about parking, we generated the shot instead, then generated it again from a reference photo until the resident looked like someone who had just come down from their apartment.
The product itself is animated rather than filmed. Parqlet picked one of four phone animations because it matched a micro-interaction already in their app, so the ad shows how the product actually feels to use.
Rounds of changes, one at a time
Most of the work after the first cut was timing. Over the second week the changes came in small, clear batches, and each one was answered with a new cut:
- The voiceover starts two seconds in, not four, so the opening doesn’t sit silent.
- A scene was lengthened so its line finishes before the next scene starts.
- Three booking rows on screen now arrive one after another, slowly enough to read.
- The music holds a steady level through the pauses, and runs two seconds past the last line.
- A pause in the middle of one line was tightened so the sentence reads as one thought.
None of these is a big change, and together they decide whether an ad feels finished. Keeping each round to a numbered list of scenes made them quick to apply and easy to check.
What Parqlet got
The final cut went out in 4K on day ten: a landscape LinkedIn ad with an AI voiceover, product animation, captions and an end card. Then one more problem turned up while Margo set up the campaign. The captions LinkedIn generates on its own spelled the company “Parklet”. We sent caption files with the right spelling in two formats; LinkedIn’s own help page says video ads take an uploaded .SRT caption file, in English.
| What | Detail |
|---|---|
| Audience | Property managers and HOA boards, on LinkedIn |
| Input | A script, reference images and app screens; no footage |
| Output | One landscape ad in 4K, with caption files |
| Voice | A licensed AI voice, chosen from samples |
| Time | Ten working days from payment to final cut |
What other startups can take from it
- Write for one buyer. Parqlet’s ad talks to the person at the front desk, not to residents or investors. Every scene answers that person’s question.
- Agree the voice before the cut. The voice sets the pace. Fixing it first saved a round of timing changes later.
- Prefer real footage, generate the gaps. Stock where it fits, generated shots only where nothing real exists, regenerated until they pass a second look.
- Ship your own captions. Automatic captions get names wrong. Upload a caption file so the company name is spelled right everywhere the ad runs.
If you are a small team with a launch coming up, this is the job our launch studio does every week: a short video from the product and whatever you already have, with a realistic AI voice if you want one. The launch video studio lists the three tiers and what each includes, and Leo for startups shows how the same footage then becomes a month of posts after launch day. We make these videos, so read this as a case study from the team that did the work.
Sources
- Parqlet, Guest parking for multifamily communities
- LinkedIn Help, Create video ads for Classic ad sets in Campaign Manager
Written by Lera Leonteva. Prices and rules change; each one is linked to its own page, read November 2026.