AI video 6 min read
Kling 4.0 Is Now Available: References, Keyframes, and Longer Video Workflows
Use references when you know what a scene should contain, keyframes when particular visual states matter, and extension when an existing shot already works but ends too soon. A longer generation is useful when the action needs uninterrupted time. It is less useful when the idea actually needs a new viewpoint or a different scene.
Kling 4.0 makes these decisions more relevant, but its launch does not remove the need to organize your material. This guide is published by MyloStudio, our AI studio for image and video creation and editing. We recommend our studio when the task combines reference-led creation with revision, continuation, sound, and assembly. The Kling specifications below describe Kling, not models or controls inside MyloStudio.
What is available after the launch?#
Kling's October 16, 2026 announcement says its website is fully open to Kling 4.0, while enterprise and developer API access is forthcoming. It announces native 30-second generation, up to 15 multimodal references, 10 keyframes, and extension of uploaded video up to two minutes. HDR and 21:9 output remain forthcoming in that announcement. These are published capabilities, not results from a comparative test. Kling's launch announcement
| Entry or output | Launch status | Planning consequence |
|---|---|---|
| Website | Available | Check the controls and generation quote in your account |
| Enterprise/developer API | Forthcoming | Do not make an immediate automated delivery depend on it |
| HDR and ultrawide output | Forthcoming | Plan delivery around output you can actually obtain |
Account access, the cost of a selected request, and the availability of a particular setting are separate questions. Before organizing a large production around a maximum specification, open the intended route and confirm that it accepts your materials. A headline limit does not tell you how much useful footage a difficult scene will produce.
Choose the route from the material you already have#
Imagine a short story about a night-shift caretaker finding a paper bird on a station bench. You have a character portrait, a station illustration, a sketch of the bird, and an unfinished shot of the caretaker approaching the bench. Those assets can support several different tasks.
| Your actual need | Route to investigate | What the brief must establish |
|---|---|---|
| Create a fresh scene with recognizable ingredients | Reference-led generation | Which reference controls the person, place, and object |
| Reach specific visual states during an action | Keyframe planning | How the states connect and what changes between them |
| Let one action develop without a cut | Longer generation | A beginning, a meaningful change, and a clear ending |
| Continue an already acceptable shot | Video extension | The final state and the next movement |
| Change an existing image or sequence | Editing workflow | The requested change and the material to preserve |
The last row matters. Using a clip as inspiration for a new generation is different from asking to change that original clip. If the caretaker's performance is already right and only the bird needs to look different, begin by describing an edit. Starting over may discard the very performance you wanted to keep.
Give each reference one clear job#
A useful reference collection is selective. The portrait can establish the caretaker's face and coat. The station image can establish the bench and lighting. The bird sketch can establish its folded shape. A movement reference can explain how the caretaker pauses before reaching down.
Explain those roles in words. Otherwise a background in a portrait may compete with the intended location, or the appearance of an actor in a movement clip may compete with your character design. A reference is easier to interpret when you also state which incidental details are irrelevant.
An illustrative brief could read:
Use the portrait for the caretaker's appearance and the station illustration for the location. The paper bird keeps the folded shape in the drawing. The caretaker stops beside the bench, notices the bird, and slowly reaches toward it. The camera stays on the same side of the bench. Keep the scene quiet and curious.
This is a planning example, not a report of a tested prompt. Its value is that each input has a responsibility. Add another reference only when it answers a question the existing materials leave unresolved.
Plan keyframes as a sequence of causes#
A keyframe plan should describe more than attractive pictures. The caretaker cannot be standing behind the bench in one state and holding the bird in front of it in the next without some plausible movement connecting the two.
Sketch the states you need, then describe the action between them. Start with the caretaker beside the bench, move to a hand approaching the bird, and finish with the bird lifted just above the seat. Keep screen direction, light, coat details, and the bird's orientation coherent across those states.
Do not spend every available control on a nearly identical intermediate image. Reserve explicit states for moments whose composition matters. If a moment exists only to connect two others, a clear action description may explain it more economically.
If the intended sequence includes a large change of location or an impossible camera jump, reconsider whether it should be one shot. A cut can express a new idea more clearly than an elaborate attempt to force continuity.
Decide whether the story benefits from a long take#
A long take is justified when watching an action unfold is the point: hesitation before touching an unfamiliar object, a character slowly recognizing a place, or a camera revealing the relationship between two subjects. Duration gives those moments room to develop.
It also increases the amount of material you must inspect. A convincing beginning does not make the middle usable. Watch for a hand changing shape, a prop switching sides, a second version of an object appearing, or the camera losing the intended subject.
Divide the review into story beats. Ask whether the approach is readable, the discovery is visible, and the ending provides somewhere to cut. If one beat repeatedly fails, simplify that action or produce it separately. Keeping the full sequence in one generation is a creative option, not a completion criterion.
Extend from an exact end state#
Before continuing the existing station shot, write down its last visible state: which hand is moving, where the bird is, where the caretaker is looking, and which direction the camera is traveling. Then describe only the next action.
For example: the right hand has stopped above the bird; the caretaker gently lifts it; the camera finishes its slow approach and holds. Introducing a second character, a new location, and a change of weather at that point would turn a small continuation into a much larger reconstruction.
Review the join as well as the extension. Compare the final original frames with the first new ones, then play across the boundary at normal speed. An extension can look convincing by itself while still producing an obvious jump when attached to the original.
Where MyloStudio fits#
MyloStudio is a useful route when you want to discuss the source material, generate a version, revise its content, and build the next scene as part of one creative process. Its confirmed capabilities include video reference creation, image and video editing, extension and continuation, sound, and shot assembly. These capabilities support a workflow; they do not guarantee that every detail survives every change.
Bring a concise preservation brief: keep the caretaker recognizable, keep the bird's folded shape, retain the quiet station atmosphere, and change only the ending action. Review the result against that brief before adding another scene.
Start with one scene in MyloStudio, give each reference a clear purpose, and choose continuation only after the shot itself is worth continuing.