The Need for Speed AI prompt

By Kemi · Updated

This is the prompt our generator sends, word for word. It is written for a video model in edit mode, with two inputs: the trend clip as [Video 1] and your car photo as [Image 1].

The prompt

Video editing task. Only replace the main foreground vehicle in [Video 1]
with the vehicle shown in [Image 1]. [Image 1] has full authority over the
vehicle: body shape, make and model, paint color, wheels, lights, trim and
any modifications, kept identical and consistent in every shot. [Video 1]
has full authority over everything else, which must stay unchanged: every
shot, cut, camera angle and camera move, the timing, the streets,
buildings, parking lots, the other cars and trucks, the bystanders, the
tire smoke, burnouts, drifting and the chase action. Same cinematic film
look, color grade, lighting and motion blur. Wherever the original hero
car appears, the vehicle from [Image 1] appears in its place, doing the
same moves at the same moments. Realistic scale and physics. No badges,
lettering, number plates, captions, subtitles or watermarks anywhere.
No dialogue and no background music.

Why each part is there

  • “Only replace the main foreground vehicle”: the clip is full of other cars and trucks. Without this, models often repaint the whole parking lot to match yours.
  • Split authority: telling the model which input controls what ([Image 1] the car, [Video 1] everything else) keeps the shots and timing from drifting.
  • “Consistent in every shot”: the clip cuts between about a dozen angles. This line reduces the car changing model between cuts.
  • No lettering or plates: AI video garbles text. Asking for none looks cleaner than half-readable badges.
  • No music: the trend’s sound is added afterwards, so generated audio would clash.

Settings that matter

SettingWhat we useWhy
ModeEdit with a reference videoCopies the clip shot by shot
Aspect ratioFollow the reference (16:9)Matches the original picture
LengthSame as the reference (≈30 s)The whole scene, in sync with the sound
Resolution720pSharp on phones; 1080p costs much more per second
AudioOffThe original sound is added afterwards

Adapting it

  • Motorbike: keep the prompt as it is, but add “ridden by a rider in a dark helmet” so the model doesn’t leave the bike riderless.
  • Lifted trucks: add “keep the lift height and tire size from [Image 1]”, because models tend to lower trucks to match the original car.
  • Other apps: some apps call the inputs @video1 and @image1 instead of [Video 1] and [Image 1]. Rename them to match.

Don’t want to tune prompts? Upload your photo here and get the finished 30-second video.

More about the trend