Videoklingkling-v3.0-pro-text-to-video

Kling 3.0

Cinematic Kling 3.0 video. Add a start frame to animate an image (image-to-video), or just prompt for text-to-video. Multi-shot prompts, native audio, and 720p / 1080p / 4K.

Open in Video Lab Browse all models~$0.94 per run

Kling guideVerified

Kuaishou's video model. The strongest motion in the open API ecosystem — fluid human bodies, action choreography, dramatic camera moves. v3.0 Pro is the flagship.

About this variant

Kling's flagship. Text-to-video, or attach a start frame to animate a still (image-to-video). 720p/1080p/4K.

Pick this for finals. Use kling-v2.5-turbo-pro-t2v for cheaper drafts.

Strengths

Best-in-class human and animal motion fidelity — runners, dancers, fighters look anatomically right.
Action choreography: martial arts, sports, combat scenes hold together over the full clip.
Camera moves: smooth dollies, crane shots, orbits without warping the subject.
Native multi-shot mode in v3 (the API exposes it as multi_prompt).
Long context for prompts — accepts dense scene descriptions without losing the lead subject.

Weaknesses

Text rendering inside images is unreliable — avoid signs / readable writing.
Hands and fingers can still distort on fast motion.
Output durations capped at ~10s per single clip (use multi-shot to extend).
Slow vs. fast tiers vary a lot in quality; Standard is often blurry compared to Pro.

Best for

Action sequences, fight choreography, sports highlights
Anime / OVA-style narrative shots (pairs well with the Anime Styles panel)
Music video moments — performers, dancers, crowd shots
Camera-heavy cinematic establishing shots

Avoid for

Tight close-ups with text (logos, signs, captions)
Long single-shot videos beyond ~10s — use multi-shot instead

Prompting tips

Lead with the subject + action, then describe the camera move, then the lighting and mood.
Be explicit about motion verbs — "runs", "jumps", "pivots", "orbits" — Kling rewards specificity.
For consistent characters across shots, repeat the same description verbatim each shot.
Use "cinematic" / "35mm anamorphic" / "golden hour" trailing tags — Kling respects them.
Avoid abstract direction ("epic", "amazing") — replace with concrete cinematography.

Parameter tips

Duration: 5s is the sweet spot. 10s costs a lot more and often introduces drift.
Aspect ratio: 16:9 is what Kling was trained on; 9:16 works but loses some quality.
Pro vs Standard: always pick Pro for finals. Use Standard only for cheap drafts.

Style packs — paste before any prompt

90s OVA cell-shaded

anime cel-shaded, 90s OVA aesthetic, hand-drawn line art, soft halation, 35mm grain

Cinematic noir

noir high contrast, chiaroscuro lighting, anamorphic lens, smoke and rain, desaturated teal-orange

Parameters

start_image_url
string
Optional. Add an image to animate it (image-to-video). Leave empty for text-to-video.
end_image_url
string
Optional. Guides the final frame (image-to-video only).
elements
array
Inject characters / objects via reference images or video clips. Reference in prompt as @Element1, @Element2.
multi_prompt
array
Sequence of shots, each with its own prompt and duration. When set, overrides the single prompt.
duration
string
No description.
3456789101112131415
default: 5
resolution
string
No description.
StandardPro4K
default: Pro
aspect_ratio
string
No description.
16:99:161:1
default: 16:9
generate_audio
boolean
Native audio with Chinese + English voice output.
default: true
shot_type
string
'intelligent' = model picks shot structure; 'customize' = you control via multi_prompt.
customizeintelligent
default: customize
negative_prompt
string
No description.
default: blur, distort, and low quality
cfg_scale
number
Higher = stricter prompt adherence.
default: 0.5
range: 0 … 1