Kling 4.0 Image to Video: Animate Photos With First and Last Frames Before Kling 5

Kling 4.0 image to video turns a still photo into a 3 to 30 second video clip: you upload a first frame, optionally add a last frame to lock the ending shot, and write a prompt describing the motion between them at 720p or 1080p. Because Kling 5 is not released as of November 2026, this is the current workflow for animating stills, and the frames you prepare now will stay useful later.

You can run this workflow right now on Kling5 AI, an independent platform not affiliated with Kuaishou or the official Kling team, using our Kling 4.0 AI video generator. Although our site name includes "Kling 5", we currently run Kling 4.0 and Kling 3.0 while waiting for a future release.

What does Kling 4.0 image to video do?

Kling 4.0 image to video takes one or two still images and generates the motion that connects them over a 3 to 30 second timeline. Kuaishou officially launched Kling 4.0 on October 15, 2026 for all users on kling.ai and opened full API access, including the /image-to-video/kling-4.0 endpoint, on October 28, 2026. Compared with Kling 3.0, which topped out at 15 seconds per generation, Kling 4.0 doubles the maximum single-pass clip length to 30 seconds and outputs in 720p or 1080p, with 4K and 10-bit HDR listed by Kling as coming soon. For a broader overview of the model beyond still-image animation, see our guide to Kling 4.0 features.

When you animate a photo with Kling, the images you upload and the text prompt you write have distinct jobs:

• Single first frame (start frame only): The image establishes the opening shot—your character, product, environment, lighting, and aspect ratio. The model holds that visual foundation at second zero and relies on your prompt to decide how the subject and camera move forward in time.
• First and last frame (start and end frames): Uploading a second image pins down the final moment of the clip as well. Kling 4.0 treats the first image as the opening frame and the second image as the closing frame, generating the transition that bridges Pose A to Pose B or Angle A to Angle B.
• The text prompt: Because the visual details already exist in your uploaded photos, the prompt does not need to re-describe the picture from scratch. Instead, the prompt tells the model what changes between the start and the end—how the subject moves, how the camera travels, the sequence of events, and what audio or text rules to follow.

In Kling's own application, image-based generation is divided into two separate entry points that cannot be combined in a single task: "Omni Reference" (which accepts images, videos, and voice) and "Keyframes" (which accepts images only for pixel-level consistency, using a first frame, first and last frames, or up to 10 keyframe images). Kling's official app also includes a lighter mode called Kling 4.0 Flash, which runs up to 20 seconds at a maximum resolution of 720p and does not currently support start-and-end frames or multiple keyframes.

On Kling5 AI, we focus specifically on the core start-and-end-frame setup inside our Kling 4.0 image-to-video tab: you provide a required first frame and an optional last frame, up to two images in total.

How do you animate a photo with Kling 4.0 on Kling5 AI?

Animating a still image on Kling5 AI follows a direct, step-by-step workflow inside the web generator, taking you from file upload to a rendered 720p or 1080p clip. Here is how the setup works on our site:

1. Open the generator: Go to the Kling 4.0 page on Kling5 AI (/kling-4), where you will see three generation modes: text-to-video, image-to-video, and reference-to-video.
2. Select the image-to-video mode: Switch to image-to-video so you can upload frames.
3. Upload your first frame: Upload your starting image as the first frame. This image is required for every image-to-video generation and sets both the opening composition and the output aspect ratio of your video.
4. Add an optional last frame: If you want the clip to land on a specific ending pose, product angle, or camera framing, upload a second image as the last frame. If you want open-ended motion driven purely by your prompt, upload only the first frame. On Kling5 AI, image-to-video takes at most two images (middle keyframes are not supported on our platform).
5. Write your motion prompt: Enter a prompt describing the action, camera movement, and sequence of events between your frames. A prompt is required on Kling5 AI and supports up to 2,500 characters.
6. Choose your duration: Pick any whole second from 3 to 30 seconds. The default setting on Kling5 AI is 5 seconds.
7. Select your resolution: Choose between 720p (the default setting) and 1080p. Notice that there is no aspect ratio selector in image-to-video mode on Kling5 AI; unlike text-to-video and reference-to-video—which let you pick 16:9, 9:16, 1:1, or 21:9—the image-to-video output automatically follows the aspect ratio of the frames you upload.
8. Check the credit cost and generate: Credit usage scales with both the duration and the resolution you select. The generator displays the exact credit cost before you start the job, and you can review plan details anytime on our pricing page.

Creators who work with long-form image-to-video clips across multiple tools sometimes also use Seedance 2.5 on AI Video Studio alongside Kling 4.0.

How to choose a good first and last frame

A smooth Kling 4.0 first and last frame transition depends on uploading two images that share the same visual world, subject identity, and framing logic. According to Kling's official start and end frames quickstart guide, the content of the two frames should be as similar as possible; if there are significant differences between the start image and the end image, the model may insert an abrupt lens or shot switch instead of a continuous motion bridge.

Practical ways to pair your start and end frames

• Keep lighting, theme, and framing consistent: Both images should look like two moments captured by the same camera in the same room or environment. If the first frame is a warm indoor close-up of a coffee mug on a walnut table and the last frame is a cool daylight wide shot of the same mug outdoors, the model has no physical path to move smoothly between them and will likely cut from one scene to the next.
• Edit your first frame to create the last frame: Kling's official start and end frames guide notes that many users generate or prepare a matching pair of similar images with an image generator before bringing them into the video workflow. If you start from an existing photograph or product shot, making a subtle edit to that exact image—such as turning a character's head, opening a box lid, or shifting a camera angle slightly while preserving the background—gives the model a clear visual path between the two states.
• Match the dimensions of both uploads: Since image-to-video follows the aspect ratio of your uploaded frames, make sure both your first frame and your last frame are cropped to the exact same ratio before uploading.

Kling's official API image requirements

If you want a clean technical baseline when exporting your source photos, Kling's official Kling 4.0 image-to-video API documentation (as of November 2026) specifies the following input image rules for the model:

• File formats: .jpg, .jpeg, or .png.
• File size: Up to 30 MB per image.
• Minimum dimensions: Both width and height must be at least 300 px.
• Aspect ratio bounds: Between 2:5 and 5:2.

Preparing your .png or .jpg frames within those official API boundaries—and cropping them to standard video ratios like 16:9 or 9:16 before uploading—helps avoid stretching or unexpected framing behavior.

How do you write a Kling image to video prompt?

A good Kling image to video prompt focuses on concrete physical movement, camera direction, event sequence, and audio boundaries rather than abstract quality buzzwords. In its Kling 4.0 API documentation, Kling advises organizing prompts around four clear elements: reference roles, overall setting, event sequence, and audio requirements.

Official prompting rules from Kling's documentation

1. Describe concrete visuals instead of abstract labels: Kling's user guide explicitly advises against vague phrases like "premium look". Instead of asking for a "cinematic luxury vibe", describe the tangible motion in the frame: condensation rolling down a glass bottle, soft studio key light shifting across brushed aluminum, or a slow camera push-in.
2. State the role of each image explicitly: In Kling's official API documentation, images can be assigned IDs and referenced in the prompt as @image_1 and @image_2. Kling notes that if you do not explicitly state the purpose of each reference in your prompt, the model will infer how to use them on its own, which may differ from your intent. Even when writing a general image-to-video prompt, clearly stating how the shot starts from the opening frame and settles into the final frame keeps the transition anchored.
3. Match your prompt timeline to your chosen duration: In the official API, if no duration parameter is sent, the model looks for a duration stated in the prompt (such as "Generate a 15-second video") and otherwise defaults to 10 seconds. On Kling5 AI, you select the exact second count (3 to 30 seconds) in the interface, so make sure the sequence of actions you write can realistically play out inside the number of seconds you pick.
4. Use the two specific negative instructions Kling recommends: According to Kling's user guide, the only negative instructions you really need in a prompt are "no subtitles" and "no background music", because the model may add subtitles or background music by default if you leave audio and text behavior unspecified.

Example prompts you can adapt

Below are three example prompt structures designed around Kling's official guidelines. These are illustrative templates for you to adapt to your own images, not outputs from hands-on test runs:

• Product rotation between start and end frames (5 to 8 seconds): "Starting from the front-facing skincare bottle in the first frame, the camera performs a slow, steady arc to the right across a dark slate surface while soft rim light reflects along the frosted glass edge, settling smoothly into the angled three-quarter product view in the last frame. Subtle water ripples move across the base. Quiet room ambience, no subtitles, no background music."
• Character motion from a single first frame (6 to 10 seconds): "In a rain-lit city street at night, the woman from the first frame slowly turns her head toward the camera, lowers her clear umbrella onto her shoulder, and takes two calm steps forward along the wet pavement as neon reflections shimmer in the puddles behind her. Gentle rain sound only, no subtitles, no background music."
• Multi-step sequence across a longer clip (12 to 15 seconds): "Generate a 15-second continuous shot in a sunlit workshop. Beginning with the closed leather journal on the oak desk in the first frame, a pair of hands enters the frame, unties the leather cord, opens the cover flat to reveal blank cream pages, and sets a brass fountain pen beside the spine to match the final frame. Soft natural room sound, no subtitles, no background music."

What duration and resolution should you pick?

Choosing between 3 and 30 seconds and between 720p and 1080p comes down to matching the complexity of your motion to your credit budget. Because Kling 4.0 lets you pick any whole-second duration from 3 to 30 seconds, you do not have to pad a simple three-second head turn into a 10-second render or cram a multi-step product reveal into five seconds.

Both in Kling's official API and on Kling5 AI, cost scales directly with clip length and output resolution. For context on how the underlying model is priced for developers, Kling's official developer pricing page (as of November 2, 2026) bills image-to-video output per second with no input video at approximately $0.157 per second for 720p and $0.235 per second for 1080p. On Kling5 AI, your generation uses platform credits that likewise increase as you raise the second count or step up from 720p to 1080p, with the exact cost displayed before you click generate.

As a practical workflow suggestion, consider drafting your prompt and checking how your first and last frames transition using a shorter duration (such as the default 5 seconds) at 720p first. Once you see that the model bridges your two frames without an unwanted shot switch and follows your camera direction, you can switch the resolution to 1080p and dial in the final second count for your finished render.

Common problems and fixes in Kling 4.0 image to video

Most issues in a Kling start and end frames workflow trace back to either a wide visual gap between the two uploaded photos or missing constraints in the prompt. If a generation does not behave as expected, check these four adjustments:

• Unexpected cut or shot switch in the middle of the clip: As noted in Kling's official start and end frames guide, significant differences between your first frame and last frame cause the model to switch lenses instead of animating a smooth transition. Fix this by using two frames with the same lighting, environment, camera focal feel, and subject scale—ideally created by lightly editing the first frame to make the second.
• The subject or background drifts away from your intent: Kling's official API guide warns that when image roles and scene details are left vague, the model infers how to use the references on its own. Fix this by explicitly stating in your prompt what the first frame represents, what must stay anchored in the setting, and the exact order of movements leading to the last frame.
• The output video has the wrong aspect ratio: On Kling5 AI, image-to-video mode does not use an aspect ratio selector; the video automatically inherits the dimensions of your uploaded frames. If your output looks cropped or vertical when you wanted horizontal, crop both your first and last frame images to your target ratio (and within Kling's 2:5 to 5:2 API ratio bounds) before uploading them.
• Unwanted captions or music appear on the clip: Kling's user guide notes that the model may add subtitles or background music by default. Add the exact negative instructions "no subtitles" and "no background music" at the end of your prompt to keep the frame clean and the audio track free of score.

Official Kling app vs Kling5 AI: what differs?

While both platforms run Kuaishou's Kling 4.0 model for 3 to 30 second image-to-video generation at 720p and 1080p, Kling's official app/API and Kling5 AI offer different interface options and keyframe limits. Here is a side-by-side look at what each environment supports as of November 2026:

If your project specifically requires chaining three to 10 middle keyframes in a single request or calling the @image_1 developer parameters directly via code, you will need Kling's official app or API. If you want a straightforward web interface for single-frame or start-and-end-frame animation—or want to switch between Kling 4.0 and Kling 3.0 in one account—Kling5 AI covers that workflow directly.

Will Kling 5 change image to video?

As of November 2026, Kling 5.0 has not been released or officially announced by Kuaishou, so there are no verified specifications for how Kling 5 image to video will work. Any websites or social posts claiming to list Kling 5.0 durations, resolutions, or frame limits right now are sharing unverified speculation. You can follow verified updates as they happen on our Kling 5.0 release date page, and Kling5 AI plans to add Kling 5.0 once Kuaishou officially releases it.

In the meantime, there is no reason to put image-to-video projects on hold. In our opinion, building a well-organized library of matching first and last frames and clear, concrete transition prompts for Kling 4.0 is likely to stay useful when a next model arrives. Pairing consistent lighting and framing across two still images is a habit that does not depend on any one model version.

When you are ready to animate a still photo or bridge a start and end frame into a 3 to 30 second video, open our Kling 4.0 AI video generator, upload your frames, choose 720p or 1080p, and generate your clip.

Sources

• Kuaishou Official Press Release: Kling AI Officially Launches Next-Generation Kling 4.0 (Oct 15, 2026)
• Kling AI Official User Guide: Kling 4.0 Model User Guide
• Kling AI Official Quickstart: Start and End Frames Guide
• Kling AI Developer Documentation: API Updates (Oct 28, 2026)
• Kling AI Developer Documentation: Kling 4.0 Image-to-Video API
• Kling AI Developer Documentation: Video API Base Pricing