Kling Image-to-Video & First-Frame Prompts (2026)
The most controllable way to make AI video in 2026 isn’t text-to-video — it’s deciding where the shot starts and where it ends. Kling’s image-to-video mode takes a first frame; its first-and-last-frame mode takes both ends of the move and generates the motion between them. For Q4 commerce content — Black Friday product spots, Christmas reveals — that control is the difference between “generate and pray” and an actual shot plan.
Most creators misuse it in one of two ways. They treat the first frame as a mood image and leave the motion to chance, or they pick first and last frames that no camera could connect — same face, different light — and get a morphing mess in between. End-frame control doesn’t remove the need for motion prompts; it raises the bar for them.
This guide covers the two-frame method: pairs that interpolate cleanly, seven shot elements for first-frame prompts, three copy-paste templates, and how to lift motion grammar from reference clips.
Two frames decide the whole clip
In first-and-last-frame mode, the generated seconds are an interpolation you are negotiating. The two frames are your terms:
- First frame = the contract. Identity, wardrobe, lighting direction, and set come from it and must stay true.
- Last frame = the destination. It must differ from the first frame in exactly one main dimension: position, scale, state, or expression. Two changes at once (moved and relit AND resized) is where morphing starts.
- The gap between them = the motion budget. A tiny gap gives micro-motion (breathing, steam, a slow push); a big gap gives action and risk. Six-second clips want medium gaps: one turn, one gesture, one reveal.
Practical rule: write the motion prompt for the gap. The frames describe states; your prompt describes how the state changes across the seconds you’re paying for.
Pick pairs that a camera could connect
Interpolation fails when the pair implies physics the model can’t fake:
| First frame | Last frame | What happens between |
|---|---|---|
| Product on table, warm light | Product lifted toward camera, same light | Clean: one hand motion, one push-in |
| Face frontal, studio light | Same face three-quarter, sunset light | Morph risk: identity + lighting both changed |
| Mug closed on counter | Mug mid-air, lid off, steam up | Clean if the hand appears in the first frame too |
| Snowy street, day | Same street, night, lit windows | Works only if architecture stays pixel-stable |
Three pairing rules. Keep the light source in the same direction across both frames. Show the agent of change — if hands lift the mug, hands enter in frame one. Move along one axis (closer, or the camera arcs, or day turns — pick one). Pairs that obey these rules need boring, mechanical motion prompts, which is what generates cleanly.
The seven shot elements for first-frame prompts
| Element | What to specify | Kling-ready phrasing |
|---|---|---|
| Subject | Identity words shared by both frames | “same woman, dark bob, gray knit sweater” |
| Framing | First-frame framing + end framing | “starts medium, ends medium close-up” |
| Camera | One move bridging the frames | “slow 15% push-in toward her” |
| Motion | What changes in the gap, timed | “0–2s: still; 2–5s: she lifts the letter; 5–6s: eyes close” |
| Light & color | Held constant across the pair | “warm lamp key from the left in both frames” |
| Timing | Duration + pacing + aspect | “6s, unhurried, 9:16” |
| Sound | Edit plan implied by the motion | “paper rustle at 2s; music holds” |
Kling-specific notes: name the direction of movement explicitly (“toward camera,” “left to right”) — with end-frame control, direction words stop the model from inventing detours. And ask for physical continuity (“the sleeve stays pushed up, the knot stays tied”): small props are the first thing to change identity between frames.
Three copy-paste templates
1. Product hand-off, vertical (9:16)
FIRST FRAME: matte steel insulated bottle standing on a linen cloth, sprig of rosemary beside it, warm side light from the left, soft shadows, photorealistic.
LAST FRAME: same bottle held up to camera by one hand, same warm light, background softly blurred.
MOTION PROMPT: same bottle, same linen set. Beats: 0–2s: static hold, condensation on the steel; 2–4s: a hand enters from frame right and lifts the bottle smoothly toward camera; 4–6s: the bottle settles in a close-up, cap angle square to camera. Camera: slow 20% push-in, no pan. Product shape, cap and lighting direction remain consistent. 6s, 9:16, one take, photorealistic.
2. Day-to-night character transition, vertical (9:16)
FIRST FRAME: young man in a black oversized coat standing on a rain-wet sidewalk, blue-hour daylight, neon storefronts unlit behind him, medium shot.
LAST FRAME: the same man in the same spot and coat, same sidewalk wet, now night — neon signs glowing behind him, reflections stretched on the ground.
MOTION PROMPT: same man, same coat, same sidewalk. Beats: 0–2s: he stands still, daylight blue; 2–4s: light shifts gradually darker as the neon behind him switches on sign by sign; 4–6s: he looks up, breath visible, reflections glowing. Camera: static medium shot with a 10% slow push. Face, wardrobe and street layout remain consistent. 6s, 9:16, cinematic realism.
3. Gift reveal between frames, landscape (16:9)
FIRST FRAME: deep-red kraft gift box with cotton ribbon, closed, on a wooden table, snowy window behind, warm lamp light, wide shot.
LAST FRAME: the same box open, ribbon loose, a folded green scarf visible inside, same table and window, same light.
MOTION PROMPT: same box, same table, same snowy window. Beats: 0–2s: static wide, snow drifting past the window; 2–4s: hands enter and pull the ribbon, the bow loosens; 4–7s: the lid lifts open, the scarf inside catches the lamp light, ribbon tail slides off the table edge. Camera: slow 15% push-in toward the box. Box color, ribbon material and window view remain consistent. 7s, 16:9, one take, warm quiet mood.
Where motion between frames comes from
Picking a gap is a cinematography decision, and cinematography is learnable from the clips in your feed. Every smooth reveal you’ve saved is data about how motion bridges two states — how fast the lift runs, when the eyes move, how long the settle takes.
With SocialToPrompt that data comes out structured. Paste a link from 25+ platforms (YouTube, TikTok, Instagram, X, Xiaohongshu, Kuaishou, Bilibili and more) or upload a file — up to 20 seconds — and get subject, framing, camera, motion, light, timing and rhythm broken out per shot. Use the first and last shots of a reference clip as your frame pair, and its middle beats as your motion prompt. A full run costs 4 credits (1 extraction + 3 generations); new accounts start with 3 credits.
Keep the grammar; redraw the frames as your own.
Pre-flight checklist
- [ ] First frame carries full identity: face, wardrobe, set, light direction
- [ ] The two frames differ in exactly one main dimension
- [ ] The light source stays in the same direction in both frames
- [ ] The agent of change (hands, wind) is visible or entering in frame one
- [ ] The motion prompt describes the gap, not the frames
- [ ] Direction words are explicit (“toward camera,” “left to right”)
- [ ] Small props are named as constant (sleeve, knot, cap angle)
- [ ] Duration, aspect ratio and a constancy clause close the prompt
Common mistakes
Frames the camera could never connect. If a real crew couldn’t shoot the move in six seconds, the model can’t either. It fakes it — badly.
Changing two variables at once. Position, lighting and scale at once is a morph generator. Change one; let the motion prompt handle the rest.
A mood image as the first frame. A pretty, ambiguous frame gives identity nothing to anchor to. The first frame is documentation.
Empty hands in frame one, product in frame two. Products don’t teleport. If hands do the work, hands start in frame.
Forgetting the last frame’s job. Without a destination, end-frame control degenerates into ordinary image-to-video.
Choose the pair, write the gap, keep the identity — and lift the in-between grammar from clips that already bridge states beautifully with SocialToPrompt.
Share Article