Beginner10 min

Generate Your First Clips in Runway and Kling

Now the fun part. You will take the prompts from your shot list and turn them into footage. We split the work: Runway for the stylized push-in shots, Kling for any shot with a person or sustained physical motion.

Step 1: Choose text-to-video or image-to-video

Text-to-video is fastest: paste the prompt, get a clip. Image-to-video gives you far more control because you first pin down the exact frame, then animate it. For your first short, use text-to-video for two shots and image-to-video for the hero shot so you feel the difference.

Step 2: Generate the hero shot from a still

Create the still in Midjourney or Flux, then upload it to Runway as the first frame. Describe only the motion you want, not the scene, because the scene is already locked in the image.

runway - image to video
first frame: hero-glass.png (from Flux)
$motion prompt: slow dolly push-in, ice settles, condensation drips
duration: 5s camera: locked seed: 48213
generating... done -> clips/runway/shot1.mp4
$

Step 3: Generate the motion shot in Kling

For the milk-pour shot, Kling holds the fluid motion more believably. Paste the prompt, set 5 seconds, and pick the higher quality mode if your credits allow. Generate two variations and keep the better one.

Always generate at least twice
AI video is a slot machine in the best sense. The same prompt gives different results each run. Two or three generations per shot is normal, and the second is often the keeper.
Kling - generation queue
Queue
shot2-milk-pour v1 [done] 5.0s
shot2-milk-pour v2 [done] 5.0s <- keep
shot3-beans v1 [generating 62%]

Step 4: Download and sort

Download every keeper into clips/runway or clips/kling. Rename them to match the shot number. A file called shot1.mp4 is findable; a file called gen_final_v3_REAL.mp4 is not.

Result: three or four short clips on disk, named by shot, ready for the timeline. Do not edit yet. Get all the raw footage first.

Hands-on tasks