Kling 4.0 features include text-to-video and image-to-video generation, Omni Reference for combining visual and voice references, up to 10 keyframes, 3–30-second videos, multi-shot direction, camera movement control, and two-channel stereo audio. Its official specifications list 720p and 1080p resolution options. References guide appearance and movement, while keyframes define specific visual moments throughout a sequence.
Last verified: 2026-10-22
How do Kling 4.0 features compare?
| Feature | Documented specification | Creative purpose |
|---|---|---|
| Reference images and elements | Up to 15 combined references, including up to 10 images, 5 videos and 7 elements | Guide subjects, appearance, action and composition |
| Keyframes | First and last frames; up to 10 keyframes | Define visual moments across the timeline |
| Duration | 3–30 seconds per generation | Plan short actions or longer continuous scenes |
| Resolution | 720p and 1080p are listed options | Choose output resolution |
| Multi-shot direction | Shot sequencing and scene transitions | Describe a sequence of different views |
| Camera control | Push-ins, pull-outs, pans, trucking, tracking and orbit shots | Direct movement through prompts or references |
| Audio | Two-channel stereo, lip synchronization and multilingual speech | Coordinate sound, dialogue and performance |
These controls address different parts of a video. A reference establishes what a subject looks like, a keyframe establishes a visual moment, and camera direction describes how the view moves. The official user guide documents the individual specifications.
How do reference images and elements work?
Omni Reference combines images, videos, subjects and voices. The 15-reference ceiling applies across the combination; the individual category limits do not add together. Up to 10 images, 5 videos and 7 elements are documented, with up to 3 video elements within the element allowance. Reference videos have a combined duration limit of 30 seconds. See the reference specifications.
Official examples include subject images, three-view character sheets, grid images, storyboards and wireframes. Video references can guide character performance, action, camera movement and narrative pacing. White-model videos and depth-map videos are also described as reference formats in the official feature announcement.
The same announcement describes creative recreation: using a reference video's pacing, composition and expressive style while replacing its products, characters or other assets. This makes the distinction between reference roles useful. A character sheet communicates appearance, while a moving reference communicates how a scene or performance develops over time.
How are keyframes different from reference images?
Kling 4.0 supports up to 10 keyframes within one generation, including first and last frames. References guide content and appearance; keyframes establish visual states along the timeline. Those states can describe a pose, an expression, a scene change or a product reveal. The model generates movement and transitions between those moments.
The official keyframe guide illustrates a 30-second product sequence with milestones at 0, 6, 12, 20 and 28 seconds: establish the setting, introduce the product, show its use, change the scene or angle, and finish on the product. This is a planning example, not a required timing pattern.
Using all 10 keyframes is unnecessary when fewer visual milestones explain the sequence. A keyframe also does not have to represent a new shot: it can define a change within the same scene. Keep character appearance and scene details consistent across references and keyframes so that the visual instructions agree.
What do multi-shot direction and camera controls change?
Multi-shot direction describes changes between shots; camera direction describes movement within them. Kling 4.0 documentation names push-ins, pull-outs, pans, trucking shots, tracking shots and orbit shots. These movements can accompany continuous action or form part of a sequence that changes viewpoint.
The official feature announcement specifies prompts of up to 8,000 tokens and describes automatic expansion and refinement of creative instructions. That length is a ceiling, not a recommended prompt size. A fixed maximum number of shots is not confirmed. The camera and prompt specifications describe improved interpretation and transitions, without establishing a numerical shot allowance.
The official feature overview connects camera direction with longer takes and multi-shot sequences. It describes keyframes as defining important moments, with prompts explaining the action and camera movement between them. For planning, distinguish the desired composition from the movement used to reach it.
What duration, resolution and audio details matter?
The documented generation range is 3–30 seconds per video. That duration can contain a continuous take or a sequence of views. The specification includes 720p and 1080p resolution options; resolution describes the output image, while duration describes how long the action unfolds.
Audio specifications include two-channel stereo and lip synchronization. Listed speech languages are English, Chinese, Japanese, Korean, Spanish, Portuguese, German, French and Hindi. The guide also names American, British and Indian English accents, alongside Chinese dialects and accents including Cantonese and Sichuan dialect. Voice references are supported. See the audio and language specifications.
What are the frequently asked questions about Kling 4.0?
How many reference images can I use?
Up to 10 images, within the combined limit of 15 reference assets. Adding videos, elements or voice references does not increase that overall ceiling.
Are keyframes the same as reference images?
No. References guide content and appearance; keyframes define visual moments. Kling 4.0 supports up to 10 keyframes, which can guide changes within a scene as well as transitions between scenes.
How long can a generated video be?
The documented duration is 3–30 seconds per generation. A longer sequence can contain multiple story moments without requiring each moment to become a separate shot.
Which resolutions are documented?
The official specification includes 720p and 1080p. These are resolution options, separate from the duration and keyframe controls used to structure a video.
Does multi-shot direction mean a fixed number of shots?
Multi-shot direction is documented, but a fixed maximum shot count is not confirmed. The limit of 10 keyframes should not be interpreted as a limit of 10 shots.
Does Kling 4.0 generate speech and other audio?
Official documentation describes two-channel stereo sound, multilingual speech, voice references and lip synchronization. The nine named speech languages are listed above. These specifications do not establish identical performance across every language or scene.
Which official sources support these details?
- Official user guide: reference limits, duration, resolution and audio
- Official keyframe guide: visual milestones and timeline planning
- Official feature overview: camera direction and shot sequencing
- Official feature announcement: reference formats and prompt specifications
Related guides: Kling 4.0 release date, video length, Omni Reference, keyframes, 4K and HDR notes, Flash, and how to use Kling 4.0.