Kling: your first clip from a still (2026)

Kling animates what you upload; it does not invent a frame from scratch as cheaply as image-to-video. This flow takes about 15 minutes and produces a short clip you can show in a pitch or a stories mock. You start from a JPG you already like — Midjourney, illustration or a real photo — and describe subject motion and, if needed, one camera move.

1. Steps 1

   Start from a still that already works

   Open the JPG you already like: product in three-quarter view, half-body portrait, landscape with one element you want to move. If it comes from Midjourney, use the grid keeper. Avoid hands in the foreground, invented type in the frame or frozen impossible actions. Do not open text-to-video yet: without an input image, the model picks composition and you pay in surprises.

   What you should see: A JPG or PNG on disk with clear frame and subject. Kling’s generator not open yet.

2. Steps 2

   Open Kling → video generator → image-to-video

   Go to klingai.com, sign in and open the video generator (Image to Video or the equivalent menu item). Drag your still or use Upload. Check aspect ratio: 16:9 for decks, 9:16 for stories, 1:1 if the JPG is already square. On the first session pick Standard or whatever the UI labels as faster for tests, not max quality yet.

   What you should see: Your image as the input frame, duration and format controls visible, empty prompt field.

3. Steps 3

   Set ~5 seconds and one gesture

   For duration, choose about 5 seconds if the UI allows (10 s on attempt 1 lengthens physics failures). One gesture: rising steam, head turning slowly, leaves moving. Do not ask for three chained actions. If there is a motion intensity control, leave it medium or low on test 1.

   What you should see: Short duration selected and a note elsewhere: «one action + optional one camera».

4. Steps 4

   Write subject + movement, not the still again

   Kling’s docs use subject + movement for image-to-video. Example: «The woman slowly turns her head toward the camera.» Product: «Steam rises gently from the mug.» Do not repeat «cinematic, golden hour, 4K» — that lives in the JPG. If you need camera, only one: «Slow push-in» or «camera holds still».

   What you should see: A 1–2 sentence motion prompt, no style paragraph. Still visible as reference beside it.

5. Steps 5

   Generate 2–3 takes and pick the keeper

   Hit Generate and wait in the queue. Repeat with the same still and gesture, or a tiny tweak («more slowly», «gentler steam»). Do not change duration or mode between the first three takes. Play each clip. Keep the take where the gesture reads on a small screen, not the shiniest or most morphing one.

   What you should see: Two or three short clips in history. One thumbnail marked as keeper.

6. Steps 6

   If morphing or hands fail, simplify

   Face morphing, extra fingers or impossible physics are still common. Do not stack adjectives: remove hands from the still, shrink the action («slight smile» instead of «talks and gestures wildly») or try a locked camera. Kling’s guides note that descriptions far from the image hurt results. Only then try a higher-quality mode with the same prompt that already read in Standard.

   What you should see: Either a cropped still without hands, a simpler prompt, or a fourth take on a higher mode with the same motion.

7. Steps 7

   Download the keeper and stop

   Download the clip you picked (the button may say Download or Export per plan). Name it by shot and action (`mug-steam-5s.mp4`). Check whether your plan adds a watermark before showing it outside an internal test. Do not cut a 60 s video in Kling — this is a prototype. Headlines, voiceover and music live in ChatGPT, ElevenLabs or your editor, not this step.

   What you should see: A short MP4 in downloads, ready for a deck or to compare with runway-primer-clip on the same still.