picturetovideo

Tool · Keyframes · Kling via API · Independent

Start and End Frame AI Video: Direct Every Beat

Start and end frame AI video lets you upload the first and last frame of a shot and have the model animate everything in between. It works today on Kling 3.0 (Studio) and Kling 2.5 Turbo (Draft) — first and last frame AI video in one render, up to 15 seconds on Kling 3.0. Kling 4.0 was announced with up to 10 keyframes per shot; that unlocks here the day its API opens, and you can plan all ten beats on the timeline now.

  • Kling 3.0 · 2.5 Turbo
  • 2 frames today
  • 10 keyframes · Kling 4.0 soon
  • From 6 cr/s
0/2500
Negative prompt
10 cr/s × 5s≈ $0.68Starter pack from $2.99
Pencil storyboard sketch of an armoured knight with a shield on a cliff, looking toward a distant castle
Start and end frame AI video illustration: a pencil storyboard of a knight on a cliff above a castle, his cape now billowing in the wind
StoryboardFinal shotKF 01 → KF 02 · 00:05
Drag the edge: sketch → shot
01The 10-slot timeline

The 10-slot keyframe AI video timeline

Drop images into slots, drag to reorder them and scrub the playhead to step through each beat. For start and end frame AI video, use the first two slots with the model you pick; switch to Chain or the Kling 4.0 preview to plan all ten.

Kling 3.0 animates one continuous shot between a start and an end frame. Slots 3–10 unlock with Kling 4.0.

Keyframe 1: a fisherman in a flat cap sorting nets in a rowboat on a grey sea, seen from above
KF 01 · 00:00.0

“The camera drifts down from overhead to the bow as he keeps working the net, grey waves rolling past”

Keyframe 01 of 2

Start + end tier · Studio (Kling 3.0)
5s

Estimate

50credits

00:05 · 2 keyframes · ≈ $0.68

  1. Start
  2. End
  3. KF 03
  4. KF 04
  5. KF 05
  6. KF 06
  7. KF 07
  8. KF 08
  9. KF 09
  10. KF 10

Drag ⠿ to reorder · drop images on a slot · arrow keys work too

02How it works

How first and last frame AI video works

A keyframe is a picture the video must hit at a given moment. In start and end frame AI video, the model invents everything in between.

Frames pin the look, prompts drive the motion

Each keyframe fixes composition, subject and light at one point in time. The text prompt describes what happens on the way to the next frame: a turn, a camera move, a change in the weather.

More frames, less drift

With only a start and an end, the model picks its own path, and long shots can wander. Extra keyframes work like waypoints. They keep a character on model and the camera on course across a longer take.

Spacing sets the pace

Frames close together produce quick changes; frames far apart give slow, smooth moves. Kling 4.0 was announced with up to 10 keyframes in a single shot of up to 30 seconds, enough to choreograph a whole scene.

Where things stand

Two-frame shots run today on Kling 3.0 (Studio) and Kling 2.5 Turbo (Draft). As of October 15, 2026 the Kling 4.0 API has not opened, so ten-frame shots on this page are planning-only until launch day. We switch them on the day the API goes live.

03Which model?

Which models do start and end frame AI video?

Lengths, resolutions and rates come straight from our model registry. Veo 3.1 Fast, Veo 3.1, Wan 2.5, Hailuo 02 and Grok Imagine start from a single image here, so they are not in this list.
Kling 3.0Kling 2.5 TurboKling 4.0
Frames per shot2 (start + end)2 (start + end)Up to 10 (announced)
Shot length3–15s5 / 10sUp to 30s (announced)
Resolution720p · 1080p720p · 1080p720p–4K (announced)
Native audioOptional—Stereo (announced)
From10 cr/s6 cr/s18 cr/s (provisional)
Best forLong, controlled movesCheap testsWhole choreographed scenes
04Kling start and end frame

Kling start and end frame in 3 steps

Make a start and end frame AI video right now with the generator at the top of this page.
  1. 01Step 1/3

    A koi fish swims through an ink painting that comes alive

    + Push in+ Orbit+ Tracking

    Upload a start frame

    This is the first picture on screen. JPG, PNG or WebP; keep the subject clear and leave some room for movement.

  2. 02Step 2/3
    Studio8s1080p16:9Sound on152 cr

    Add the end frame

    Same subject, same lighting, a new position or angle. The closer the two frames are in look, the cleaner the path between them.

  3. 03Step 3/3
    Ready · 00:08Download MP4

    Describe the move and render

    Say how to get there, for example “the kite lifts off and the camera tilts up”. Pick a tier and a length; Studio (Kling 3.0) costs 10 credits per second at 720p.

05Keyframe chaining (beta)

Keyframe chaining (beta)

Need more than two beats before 4.0 arrives? Split the shot into start and end frame AI video segments that share a frame, render each one, then join them. Here is how that compares with native 4.0 keyframes.
Chain on Kling 3.0Native Kling 4.0
AvailableToday, as a beta workflowWhen the 4.0 API opens
KeyframesUp to 10, as 9 linked segmentsUp to 10 in one shot (announced)
Length3–15s per segmentUp to 30s per shot (announced)
MotionEach segment plans its own pathOne continuous path
CostSum of segments · 10 cr/s at 720p18 cr/s at 720p (provisional)
Best forStoryboards, product turns, simple pathsLong, choreographed takes
06Storyboard tips

Tips for cleaner start and end frame AI video

01

Lock the look

Make every keyframe from the same character sheet, outfit and lighting setup. A change of wardrobe or time of day between frames reads as a glitch.

02

Move one thing at a time

Between two frames, change either the subject’s pose or the camera angle, not both, and never the whole scene at once.

03

Keep the horizon steady

A level horizon and the same lens length across frames make the camera path believable instead of floaty.

04

Space beats evenly

Uneven spacing makes speed jump. If one beat needs more time, give it a longer segment rather than crowding the others.

05

Write the in-between

Describe the journey, not the destination: “she turns and walks to the window” works better than “a woman by a window”.

06

Sketch first

Rough thumbnails are enough to plan the beats. Turn them into finished frames with an image model, then drop them on the timeline.

07Cost

What keyframe AI video costs

Credits are charged per generated second, and a chained shot costs the sum of its segments. Kling 4.0 rows show our provisional launch price, which we will only ever adjust downward.
ModelResolutionCredits / secWith audioTypical clip
Kling 3.0720p10145s · 50 cr ≈ $0.68
Kling 3.01080p13195s · 65 cr ≈ $0.88
Kling 3.04K47—5s · 235 cr ≈ $3.19
Kling 2.5 Turbo720p6—5s · 30 cr ≈ $0.41
Kling 2.5 Turbo1080p6—5s · 30 cr ≈ $0.41
Kling 4.0launch price720p18275s · 90 cr ≈ $1.22
Kling 4.0launch price1080p24355s · 120 cr ≈ $1.63
Kling 4.0launch price4K87—5s · 435 cr ≈ $5.91

Example: a 5-second start/end shot on Kling 3.0 at 720p costs 50 credits (≈ $0.68 on the Pro plan). Chaining four keyframes as three 5-second segments costs 150 credits.

FAQ

Keyframe AI video FAQ

Still curious? Email support@picturetovideo.org — a human replies within a day.

01What is start and end frame AI video?
It is image to video with two anchors instead of one. You upload the picture the shot should open on and the picture it should finish on, add a short prompt about the movement, and the model renders a continuous clip that travels from one to the other. It gives you far more control over where a shot lands than a single photo does.
02Which models support first and last frame AI video here?
Today: Kling 3.0 (Studio) and Kling 2.5 Turbo (Draft). They all accept a start and an end image in one render, at the lengths and resolutions shown in the comparison table. Veo 3.1 Fast, Veo 3.1, Wan 2.5, Hailuo 02 and Grok Imagine animate from a single image on this site. Kling 4.0 adds up to ten keyframes once its API opens.
03Can I use 10 keyframes right now?
Not natively. As of October 15, 2026, the Kling 4.0 API is not open and no provider offers it. You can plan all ten slots in the timeline editor today, render two-frame shots on the models above, or chain several start/end segments as a beta workaround. Native ten-frame shots switch on the day the API opens.
04How is chaining different from real Kling 4.0 keyframes?
Chaining renders each pair of neighbouring keyframes as its own clip. Because neighbours share a frame, the joins line up exactly, but each segment plans its own motion, so speed or camera direction can change at a join. Native 4.0 keyframes are expected to solve the whole path in a single pass, which keeps the motion continuous.
05Do my start and end frames need the same size?
They should. Use one aspect ratio and resolution for every keyframe, because the output follows the shape of your first frame. Mixing a 16:9 start with a 9:16 end forces the model to crop or stretch, which shows up as warping. Export every frame from the same canvas size before uploading.
06Can I use start and end frames for an AI video transition?
Yes. Use the last frame of one scene as the start and the first frame of the next scene as the end, then describe the move between them, such as “the camera pushes through the window into the garden”. The start and end frame AI video becomes the bridge. Transitions look smoothest when the two frames share colours, light or a shape the camera can travel through.
07How different can the first and last frame be?
Closer than you might think. The model has to invent a believable path, so keep the same subject, outfit, location and light, and change one main thing — a pose, a camera angle or a position in the frame. If the end frame needs a different place or time of day, put a middle keyframe between them and chain two segments.
08How much will Kling 4.0 keyframes cost?
Kuaishou has not published Kling 4.0 API prices. Our provisional launch rate is 18 credits per second at 720p and 24 at 1080p, which is what the estimates on this page use. We recalculate within a day of official pricing and only adjust downward. Two-frame shots on Kling 3.0 cost 10 credits per second at 720p today.
09Is this keyframe tool part of Kling AI?
No. PictureToVideo is an independent site, not affiliated with or endorsed by Kuaishou or Kling AI. We run Kling models through their API; the generator labels each render with its quality tier (Studio, Standard, Draft), and the comparison above lists the model behind each tier. Kling 4.0 keyframes will only appear as a working option once they are genuinely available through the API.
Keep exploring

Related tools & models

AI Video Extender
AI Video ExtenderContinue a finished shot from its last frame.
Kling 4.0
Kling 4.0Live 4.0 status, the waitlist and what you can make today.
Kling 4.0 vs 3.0
Kling 4.0 vs 3.0Specs side by side, and whether it is worth waiting.
Kling 3.0
Kling 3.0The model behind most start and end frame shots here.

Direct the shot beat by beat

Make a start and end frame AI video today. All ten keyframes unlock the day the Kling 4.0 API opens.