Kling 3.0

Overview

kling-3.0-omni/text-to-video is the pure text-to-video pipeline in the Kling 3.0 Omni series, called through the unified HiAPI.ai interface. It produces a finished clip directly from a prompt, supports durations from 3 to 15 seconds, 720p/1080p/4k resolutions and landscape, portrait and square aspect ratios, and can enable native audio and automatic multi-shot storyboarding. It suits developers and content teams that need to quickly validate creative shots and batch-produce short clips.

Highlights

• Prompt-driven finished video — just a single prompt (up to 3072 characters), no reference material needed, to generate a complete video clip.
• Optional native synchronized audio — with the audio parameter enabled, video and sound are generated in the same pass, skipping post-production dubbing and sound-effect alignment.
• Automatic multi-shot storyboarding — prefer_multi_shots lets the model plan multiple shots on its own from the prompt, delivering a segment with cuts in a single request.
• Reusable subjects and @ references — register up to 7 subjects, each defined by 2 to 4 images; reference them in the prompt as @name to keep that subject consistent across multiple shots.

Capabilities

Text-to-video generation: returns a video file directly after a prompt is submitted; duration is adjustable from 3 to 15 seconds to suit clips of different pacing.

Resolution and aspect ratio control: offers three resolution tiers — 720p, 1080p and 4k — and three aspect ratios — 16:9, 9:16 and 1:1 — so a single request can produce footage suited to landscape placement, portrait feeds or square assets.

Native audio switch: audio is off by default; when enabled, the video is generated together with synchronized sound, so dialogue or ambience clips need no separate audio-track assembly.

Subject registration and referencing: pass subject names, descriptions and 2 to 4 reference images via elements, then refer to them in the prompt as @name so the same character or object keeps a consistent appearance across multiple shots.

Intelligent multi-shot planning: with prefer_multi_shots enabled, the model breaks the prompt into multiple shots and organizes the edit itself, reducing the work of generating each shot separately and stitching them together.

Use cases

• Batch production of social short videos by content teams — input product selling points or topic descriptions and generate clips at 9:16 portrait, ready for feed and short-video platform placement.
• Early storyboard previews for film and animation — describe shots and staging in text, enable multi-shot storyboarding, and quickly obtain visual segments for storyboard discussions and shot design.
• Creative validation for ads and brand films — generate concept clips at 16:9 and high resolution before formal production to confirm the creative direction with clients.
• Dialogue-driven character clips with lines — register character subjects and reference them in the prompt, enable native audio, and produce character clips with synchronized sound for story samples.

Limitations and selection advice

A single generation is capped at 15 seconds; longer narratives need to be generated in segments and then stitched together. elements only provides subject appearance references; the prompt must explicitly reference them as @name for them to take effect, and unregistered subjects are not guaranteed to stay consistent. Raising the resolution to 4k significantly increases generation and transfer overhead. This pipeline currently accepts text input only; if you need to start from images or existing video, use the image-to-video, reference-to-video or video-to-video (transformation) pipelines in the same series instead. Cost is billed by the number of seconds of output video, so long durations combined with high resolution require estimating usage in advance.