Silent AI video has a tell. Even a beautifully rendered clip feels like a mood board until someone drops sound under it, and that sound step used to be a separate job with separate tools. Kuaishou's latest Kling model folds it into generation. You describe the picture and the soundtrack in one prompt, and the clip comes back with a stereo track already in sync.
That is the headline, but not the only change. The model also speaks written dialogue with matching lip movement, accepts any length from 3 to 15 seconds, and has a multi-shot mode that strings up to six shots into a single clip. Here is how each piece works and where it still needs help.
Sound that follows the prompt
The native audio is not a generic ambience bed. It responds to what you name. End your prompt with a "Sound:" line listing specific sources, like an engine roar echoing off tunnel walls, water hissing off a tire or a rising gear shift whine, and those are what you hear. Add "no music" when you want a clean effects track to score later. Leave music unspecified and you are handing that decision to the model, which is rarely what an editor wants.
Sound is on by default. Turning it off drops the per second price, which is handy when a clip will sit under a voiceover anyway.
Writing dialogue it will actually speak
Dialogue is where formatting matters most. The pattern that works puts the speaker and voice in square brackets, then one short line in quotes:
[Old fisherman, hoarse whisper]: "Tide's turning. Pull the nets."
Then close the prompt with "No subtitles, no captions, no on-screen text." That line keeps the words spoken rather than printed across the frame. Stick to one speaker and one short line per clip. Long speeches and back and forth exchanges are still better built from separate generations.
Multi-shot scenes of up to six shots
Switch on multi-shot and you write up to six shots, each with its own prompt, and Kling 3.0 renders them as one clip priced on its total length. That differs from models that parse "Shot 1, Shot 2" out of a single paragraph. Separate prompts per shot make framing easier to control and keep each beat clean.
A practical use: a 12 second product spot built as four 3 second shots. A wide shot of a kitchen at dawn, a close-up of a kettle starting to steam, a macro of tea leaves hitting water, and a medium shot of a hand lifting the cup. One action and one camera move per shot. The whole thing costs the same as a single 12 second clip.
Standard vs Pro, in plain numbers
Both tiers share the same prompt box, lengths and features, including image to video from a start frame with an optional end frame. The difference is output and price.
Standard renders 720p: 1280 by 720 widescreen, 960 by 960 square or 720 by 1280 vertical, at 3 credits a second or 4 with sound. Pro renders 1080p: 1920 by 1080, 1440 by 1440 or 1080 by 1920, at 4 credits a second or 5.5 with sound. Both run at 24 fps.
So a 5 second Pro clip with audio is 27.5 credits, a 10 second one is 55, and a 3 second Standard clip without sound is 9. The sensible approach is to rough out ideas on Standard, where a failed experiment is cheap, then move the winning prompt to Pro for anything you publish. A vertical ad at 1080 by 1920 is a Pro job. A quick reaction clip for a story is a Standard job.
Length is the quieter upgrade. The older Kling 2.6 Pro, still in the same tool, only offers 5 or 10 seconds. The new model takes any whole second from 3 to 15, so a 7 second cutaway costs exactly 7 seconds of credits instead of being rounded up to 10.
What it does best
This model is strongest with physical motion that has consequences: a motorcycle carving a rooster tail of spray through a flooded underpass, an FPV drone plunging down a waterfall and pulling out over the river, a pancake tumbling out of a pan. Name the physics you want to see and it tends to render it. Stylized looks like claymation also hold up well.
A prompt worth trying: "Low tracking shot beside a skateboarder bombing a wet hill at dusk, wheels throwing a fine mist, streetlights streaking past, 24mm lens, slight shake. Sound: wheels on wet asphalt, wind, a distant car horn. No music."
Honest limitations
Fifteen seconds is the ceiling per generation, so longer pieces still need an edit. Dialogue works best short, and multi-speaker exchanges are hit and miss. Pro's audio pricing also adds up on long clips: 15 seconds with sound is 82.5 credits, so keep audio off while you iterate on the picture.
On PixelDojo the model sits next to Kling 2.6 Pro in the same tool, which makes side by side comparison easy. For shots where a person and a place carry the story, and you want sound finished in the same pass, it is one of the most complete options available right now.