Kling: Video v3.0 Pro

Kling VIDEO 3.0 Pro is the professional profile of Kuaishou’s VIDEO 3.0 model. Its documented Pro mode produces 1080p video from a text brief or an image-led scene, with clips from three to fifteen seconds. The 3.0 edition combines native audio, subject-consistency controls and multi-shot direction, giving a short production a clear way to specify appearance, action, camera coverage and sound.

Begin with the scene’s main action. In a text-led shot, describe the subject, environment, light and the viewpoint that makes the action readable. In an image-led shot, let the opening image establish the appearance and composition, then use the text for what changes over time. For an original product concept, a prepared image can show the object on a studio surface while the prompt describes a slow reveal of its material and shape.

Multi-shot direction offers automatic coverage or a custom sequence. Automatic multi-shot generation plans changes in framing and camera angle from the brief; a scene that fits one continuous view can still remain a single shot. A custom storyboard lets you define the content and duration of each shot. The producer’s custom-shot profile permits one to six shots, with at most 512 characters per shot and a minimum of one second for each; their durations must sum to the total clip duration.

For a short narrative, describe an establishing view, the main action and a concluding detail as separate shots. State the camera treatment and what the audience should notice in each one. Keep a recurring subject’s appearance consistent in the brief, while changing the action and coverage deliberately. The single-prompt field in the maker’s reference accepts up to 2,500 characters; a custom sequence uses its separate per-shot descriptions.

Subject references can support an image-led scene when the character, product or setting should remain recognisable. Choose clear views that communicate the features you need to retain, and name each referenced subject consistently. The documented image-led profile allows up to three prepared elements. Keep reference roles focused rather than combining several assets that give conflicting instructions about the subject’s appearance.

Start and end frames provide visual boundaries for an image-conditioned transition. Prepare frames that agree on the subject and intended scene, then describe how the movement connects them. The maker’s supported frame input uses JPG, JPEG or PNG images, with files no larger than 10 MB, dimensions of at least 300 pixels and a start-image aspect ratio between 1:2.5 and 2.5:1. A frame-controlled task and a custom storyboard have their own input handling, so organise the material for the selected workflow before writing the full brief.

Native audio can combine spoken lines, atmosphere and the visual performance. Audio is off by default in the maker’s reference and is enabled for an audiovisual scene. The 3.0 language profile includes Chinese, English, Japanese, Korean and Spanish, along with code-switching and supported dialect or accent treatments. The maker’s guide says dialogue entered in another language is translated into English. Assign each line to the intended subject and distinguish dialogue from background sound, so the audio brief explains who speaks and what surrounds the action.

An original two-character brief can identify Speaker A and Speaker B, give each a short line and describe the visible response between them. Keep the dialogue proportionate to the clip length and identify the speaker when the view changes. Voice direction and a character reference serve different purposes; the brief should state the intended performance rather than assuming a reference image contains the words to be spoken.

VIDEO 3.0 also improves the handling of visible lettering in an image-led scene. If a sign, label or caption matters, identify it in the brief and preserve its reference appearance. Review the completed clip for spelling, readability and movement of that lettering, particularly when it carries a product or campaign message. A model’s lettering feature guides generation rather than replacing a final visual check.

The Pro profile is 1080p; the Standard profile is 720p. The producer’s separately selected 4K mode is another output profile, so Pro should not be treated as a synonym for every higher-resolution setting. For text-led generation, the documented framing choices include 16:9, 9:16 and 1:1. Choose the frame for the intended placement and keep the composition readable within it.

Generation runs as a task whose successful result exposes the completed video artifact and actual duration. The maker’s result documentation says generated videos are deleted after 30 days, so save the completed clip promptly. Preserve the video with its brief and reference assets, then review subject continuity, requested shot order, camera movement, dialogue and lettering. Use the selected clip as a cinematic concept, product scene or short narrative, and make a focused revision to the part that missed the intended result.