Kling 3.0 Review: Motion Control and Multi-Shot, Filtered at L2

We don't publish our content-ladder runs; what follows is the everyday register, taken from the published, safe-for-work examples on the Kling 3.0 model page.

A studio dance, recast. A driving clip of side steps, a deep squat with arms spread and both arms raised came back matched beat for beat on a new character, who kept her own face, satin suit and studio background instead of the driving clip's performer.

A turn toward camera at dusk. From a yacht-deck opening frame, "one continuous shot with no cuts" plus a light description kept the dusk color and framing steady through a full turn toward camera and a spoken line.

Lightning over a forecourt. A rain-and-lightning prompt added only the flash, a step off a truck and a walk into the downpour — the wardrobe and lighting held, but a truck's position relative to the building didn't match between the opening frame and a later angle in the same clip.

A rooftop toast. A handheld shot wove through a dancing crowd and ended with the whole group shouting one line together, timed against music and clinking bottles — naming a single shared line in the prompt kept the dialogue from overlapping.

On image-to-video, say explicitly whether you want a single continuous shot or a cut sequence — the model will add a cut on its own if you leave it ambiguous. On Motion Control, the driving clip sets the timing and the performance; the character photo sets who's performing it.

1. Locking a single shot on image-to-video.

One continuous shot with no cuts, the golden-pink dusk light staying the same throughout. He turns from the rail toward the camera, pushes his wet hair back with one hand and smiles, then says, "Jump in, the water's perfect."

Run this prompt on Kling 3.0 →

2. A single shared line for a group scene.

A summer rooftop party in Brooklyn at blue hour, string lights and the Manhattan skyline behind. A group of friends dance and laugh, then everyone shouts one line together: "To us!" Their glasses clink on the beat.

Run this prompt on Kling 3.0 →

Weak: leaving the shot count to the model, even with shot_type: single, is where we saw unasked-for cuts creep in. Strong: spell out "one continuous shot with no cuts" the way prompt 1 does when you need the take to stay in one piece; save shot_type: multi and a numbered shot list for when you want the cuts yourself.

Kling reviews every request itself, at three points: the prompt text, any uploaded image or driving clip, and the finished clip. That held across all three Kling 3.0 tasks — the same outcomes came back on every route we tested, so this is Kling's own moderation rather than something a different endpoint would change.

On 28 September 2026 we ran our five-rung ladder against image-to-video and text-to-video (Motion Control wasn't part of this round). L1 Suggestive (lingerie, a slipping strap) rendered on every task tested; L2 Partial was refused every time we sent it, including six further L2 probes across six Kling 3.0 and O3 endpoints the same day. L3 Strong and L4 Intense (an intimate scene between two adults) were refused too, mostly at the prompt or upload stage within seconds to about half a minute; on one route the L3 text-to-video request came back refused only after about two minutes. The L5 Extreme restraint-and-impact probe was refused after rendering. The L5 Extreme gore probe split by task: it rendered on text-to-video and was refused on image-to-video.

Refused requests cost nothing — every refusal in our test set, including ones that only came back refused after the clip had fully rendered, was free. There's no quiet softening either: in every run the clip either came back as written or the job failed outright, with clothing never added back and the scene never swapped for a safer one.

The page label reflects this: image-to-video and text-to-video both carry a filtered policy tier, and we add no content filter of our own on top of it.

Not tested yet: Motion Control against our content ladder; whether a longer, more specific prompt moves the L4 Intense refusal point; scenes with more than two people at the harder rungs.

Adults only, in every prompt, upload and output. Never depict a real, identifiable person without their consent, and make sure you hold the rights to any photo or clip you upload. For content past lingerie level, use one of our Spicy families instead — Wan 2.2 Spicy and Seedance 2.5 Spicy are two of the video ones — this endpoint refuses that far past L1.

Kling 3.0 comes up by name when developers compare closed-source video tools. In an April 2026 Hacker News thread, one commenter wrote: "The models most professionals use are Kling o3, Kling 3.0 and, more recently, Seedance 2.0". A separate May 2026 thread, summarizing AI-video subreddit opinion, grouped Kling 3 with Seedance 2.0 among the closed-source video models that community reached for.

Public blind-vote boards place it a little higher than O3 on the same tests. On Artificial Analysis's image-to-video leaderboard (28 September 2026), Kling 3.0's 720p Standard tier placed 23rd of 82 tracked models; on arena.ai's image-to-video board (21 September 2026), its 1080p Pro tier placed 19th of 48 on more than 200,000 votes.

• Need reference photos cast into a shot, or an existing clip edited? Kling O3 swaps Motion Control for reference-to-video and video-edit, at the same core prices as 3.0's image and text tasks. Our Kling O3 review.
• Don't need Motion Control at all? Kling 2.6 covers image and text-to-video with native audio from $0.069 a second.
• Need a single continuous take longer than 15 seconds? Wan 3.0 runs up to 30 seconds with multi-shot direction, from $0.045 a second at 480p.
• Want dialogue-heavy scenes to be the main focus? Seedance 2.5 is priced higher but built around lip-synced speech.