Back to blog

How to post a Roadster AI video on TikTok, Reels and Shorts

Oct 17, 2026Roadster AI Video TeamRoadster AI Video Team

Export settings, hook structure, AI labels and sound choices for posting a generated car stunt video without losing quality or reach.

How to post a Roadster AI video on TikTok, Reels and Shorts

A vertical AI car clip has about one second to prove it is worth watching. The algorithm does not care how long the render took or how much the credits cost. It cares whether a viewer stops scrolling, watches to the end, and comments. This guide covers the export settings, the first-second structure and the disclosure rules that keep a synthetic clip from being taken down.

Export in 9:16, then do not crop again

Generate at 9:16. TikTok, Reels and Shorts are all vertical-first; a landscape clip with black bars loses a third of the screen. If you generated 16:9 and need a vertical version, reframe in an editor rather than uploading the landscape file — the platform's own crop is centred and will usually cut the car.

Keep the MP4 as downloaded. The studio does not add a watermark, so there is nothing to crop out. Re-encoding a clean file two or three times adds softness; upload the original if the platform accepts it.

The first second

The most reliable opening for this trend is the impossible action already in progress. Do not start with a logo, a title card or a slow pan across a parked car. Start with the car in the air, or at the instant the thrusters fire.

A three-beat structure works:

  1. 0–1s: the car is already doing the impossible thing. A caption names the joke: "my 2012 hatchback finally got the update".
  2. 1–5s: the action resolves. The car lands, the smoke clears, the camera settles.
  3. 5–12s: a reaction or a punchline. The driver's face, a second angle, or a text overlay that undercuts the stunt.

Twelve seconds is the default clip length for a reason. It is long enough to tell a small story and short enough to loop. A loop is the cheapest way to raise watch time: if the last frame matches the first, the viewer watches twice before noticing.

Sound

The model does not produce a soundtrack. Add one in the editor. Three options, in order of effort:

  • Use the platform's trending sound. The algorithm already knows it performs. Pick one that fits the tempo of the action.
  • Use an engine or whoosh effect. A real engine note under a fake car reads as a joke; silence reads as a bug.
  • Record a reaction. A friend saying "what?" over the clip is often funnier than any produced sound.

Do not ask the model to generate speech or a caption. Text in AI video is unreliable, and a garbled caption destroys the illusion faster than a wobbly wheel.

The caption and the AI label

Every major platform now expects synthetic media to be disclosed. Use the platform's own "AI-generated" toggle where it exists, and add a plain-language line in the caption: "AI-generated, not real footage."

That line does three things. It keeps you inside the platform's rules. It keeps you inside our terms. And it makes the joke land: the clip is funny because everyone knows it is fake.

Avoid framing that implies a real event: "caught on camera", "Tesla's new feature", "local man flies over traffic". A fictional label is not a licence to fabricate news.

Caption formulas that fit the format

  • "POV: you finally found the boost button."
  • "My car has one feature and it is this."
  • "Day 1 of owning a flying [car model]."
  • "The commute if I had 30 more credits."
  • "Rate the landing."

Keep the caption under two lines. The video is the hook; the caption is the punchline.

What not to do

  • Do not repost someone else's clip. The trend is built on your own car, and reposting is both a copyright problem and a boring post.
  • Do not remove the AI label to chase engagement. A takedown or a fact-check label costs more reach than the disclosure ever will.
  • Do not upload a 1080p file through three editors. Each re-encode softens the image; keep one clean master.
  • Do not post the same clip to all three platforms at once with the same caption. Each platform rewards a native tone; a TikTok caption reads as spam on LinkedIn.

Test two hooks, not two videos

The cheapest improvement is to keep the clip and change the first second. Export one version that opens on the car already in the air, and one that opens on the moment the thrusters fire. Post them a day apart with the same caption family, then compare three-second retention, not likes. A hook that holds three seconds is doing more for distribution than a prettier landing frame. Retire the weaker hook and reuse the winner on the next render. This keeps production cost flat while the account learns what its audience actually stops for.

A repeatable posting loop

  1. Generate two variants of the same scene at 720p. Pick the one with the cleaner body shape.
  2. Re-render the winner at 1080p if you are posting to a platform that rewards sharpness.
  3. Edit in the sound and the caption. Add the AI label.
  4. Post at a consistent time for two weeks, then compare retention, not likes.
  5. Keep the prompts that produced the best retention; retire the rest.

The studio is below if you want to make the next clip now.

Resolution
Length
Format
This video: 35 credits · 720p · 12s

This clip: 35 credits · 10 clips a month for $39.90 (Pro), or 102 a year for $319

Frequently asked

What aspect ratio should I export? 9:16 for TikTok, Reels and Shorts. 16:9 for YouTube and X. 1:1 for feed posts.

Does the video have sound? No. Add music or effects in the editor.

Will the platform remove an AI clip? Not if it is labelled and does not impersonate a real event or person. Unlabelled realistic fakes are the ones that get removed.

Can I monetise the clip? You own the render and can use it commercially, subject to the platform's synthetic-media rules and our terms.