How to make a TikTok ad for a local restaurant
Give Vizard Agent the script or just the offer and it builds the ad. Vizard Agent sources food footage for every beat, records an energetic read, assembles the audio with ducking automated before the picture is cut, and checks the badges on rendered frames until they sit where they can be read.
What is the short version?
A restaurant ad on a short-video feed has about two seconds to work. That means an extreme close-up on the food first, a voice with energy, text that survives being watched at arm's length, and a local call to action at the end.
- Go to Vizard Agent and give it the script or the offer.
- Say the goal — comments, shares, people through the door.
- Say the length and the shape.
What do you need before you start?
The offer and the place. Vizard Agent sources the food footage and writes to your script, so you need no filming at all, and the details worth supplying are the ones only you know: the deal, the hours, and what makes people come back.
- The script, or the offer. Either works as a starting point.
- The restaurant's name. For the badges and the closing card.
- The goal. It shapes the call to action.
- The length. Thirty seconds is the format.
- Your own footage. Optional; Vizard Agent sources it otherwise.
What do you type into Vizard Agent?
State the goal alongside the script. Vizard Agent uses "get comments, shares and local customers" to shape the call to action and the pacing, which produces a different ad from one built to look appetising and stop there.
Prompt
Variants worth knowing:
- A cheese-pull opening. It is a cliché because it works.
- Bold badge overlays. Price, hours, location, in high-contrast type.
- A specific dish. Naming it narrows the footage search considerably.
What does Vizard Agent actually do?
Here is the order Vizard Agent worked in on a real restaurant ad. The notable choice is at step seven: the whole audio track was built and balanced before the picture was cut, so the visuals could be timed to a finished soundtrack.
- Checks the tools and the product-ad guidance, then searches for energetic commercial voices.
- Searches stock libraries for footage covering every scene in the script.
- Searches for energetic beats and checks what sound effects are available.
- Generates voiceover test samples, checks their durations, then records all the lines.
- Builds contact sheets of the raw footage and inspects the cheese pull, the dough, the toppings and the oven shots.
- Extracts detailed frame samples at specific timestamps to choose precise cut points.
- Assembles the master audio with volume automation and ducking, then verifies the loudness balance across every slice.
- Extracts word timestamps from the finished narration and generates animated captions to match.
- Renders the graphics, cuts the picture frame-exactly to the audio, then regenerates the captions bolder and moves the badges three times, checking each version on rendered frames.
Step 9's repetition is not indecision. A badge that looks placed correctly in a layout can sit over a bright cheese pull or under the platform's own interface, and the only reliable check is a rendered frame.
What does the result look like?
From the run this page is written from, probed on the delivered file: 1080x1920, H.264, 30fps, 30.0 seconds, AAC audio. Vertical, exactly thirty seconds, sourced food footage cut frame-exactly to an energetic narration, with bold animated captions, badge overlays and a local call to action.
Landing on exactly thirty seconds comes from cutting the picture to a finished audio track. The narration sets the length, the cuts are made frame-exactly against it, and nothing has to be stretched or trimmed at the end to fit.
When does this not work well?
Stock food footage sells an idea of a meal rather than anything that came out of your kitchen, and for a local business that distinction carries weight. Vizard Agent makes the sourced footage look excellent, and there is still a gap between the ad and the plate that regulars will notice.
- Sourced footage is not your food. For a local business that gap matters more than for a brand.
- Claims about the offer are yours. Vizard Agent puts up what you supply.
- Platform ad rules apply. Especially around health, alcohol and pricing claims.
- Thirty seconds sells one thing. The dish or the deal, not the whole menu.
- Local relevance has to come from you. The neighbourhood, the hours, the reason to walk in.
How do you fix a result that came back wrong?
Name the badge or the shot. Vizard Agent keeps every footage candidate, the frame samples, the master audio with its ducking and the badge versions, so a repositioned overlay or a swapped shot is a re-render rather than a rebuild.
- "That is not how our pizza looks." Swapped, or replaced with your own footage.
- "The price badge is hidden." Repositioned and verified on a frame.
- "The voice is too shouty." Another candidate, tested before the full read.
How does Vizard Agent compare to doing it yourself?
By hand this is filming your own food badly under restaurant lighting, or licensing stock that looks nothing like your kitchen, then fighting a template to make the price legible. The badges are the part that eats the afternoon, because you cannot tell whether they work until you watch the export on a phone.
| By hand | Vizard Agent | |
|---|---|---|
| The footage | Film it, or take what fits | Searched per scene, frames sampled for cuts |
| Order of work | Cut picture, add audio | Audio built and ducked first, picture cut to it |
| Badge placement | Judge in the layout | Checked on rendered frames, moved three times |
| Caption weight | One preset | Regenerated bolder after seeing it on frame |
Common questions
Why do the badges get moved so many times? Because where a badge sits is only testable on a rendered frame. Vizard Agent moved them three times here, checking each version against the actual picture, since a price overlay that lands on a bright cheese pull or under the platform's own buttons is simply not read.
Do I need to film anything? No. Vizard Agent searches the stock libraries and sources footage covering every beat of the script.
Can I use my own food photos or clips? Yes, and for a local business it is worth it. Give them to Vizard Agent and it builds around them, since sourced footage is not your kitchen.
How long should a restaurant ad be? Thirty seconds. One dish or one deal, told cleanly.
Will the price be readable? Vizard Agent checks the badges on rendered frames and repositions them until they are.
Why build the audio first? Because the narration decides the length. Cutting picture to a finished, balanced audio track means the ad lands on thirty seconds exactly without stretching a shot to fill a gap.
Can it write the script? Yes. Give Vizard Agent the offer and it writes the hook, the body and the call to action from it.
Does it add captions? Yes, animated and bold, generated from the narration's own word timings.
Does Vizard Agent check the finished ad? Yes. It reviews frame grids several times over and ran a full analysis on the final render.