Kling 3.0

Positioning Overview

kling-3.0-omni/image-to-video is the Omni version for image-to-video in the kling-3.0 series, invoked through the unified HiAPI.ai interface. Developers input one or more frame images together with a text prompt to generate videos of 3–15 seconds at 720p to 4k, with optional native audio, intelligent multi-shot, and reusable subjects. It suits developers who need to quickly turn static images into dynamic videos for scenarios such as e-commerce showcases, social media clips, and storyboard previews, while requiring element consistency and controllable duration and image quality.

Core Highlights

• First/last frame dual-image control — supports 1 image as the first frame, or 2 images specifying the first and last frames respectively, giving the generation process a clear start and end point.
• Reusable subjects maintain consistency — define up to 3 subjects, each described with 2–4 images and referenced in the prompt as @name, reducing subject drift across shots.
• Optional native audio output — when audio is enabled, the video and synchronized sound are generated together, eliminating the separate post-production step of voiceover and audio-video alignment.
• Adjustable duration and resolution — output duration covers 3–15 seconds, and resolution can be set to 720p, 1080p, or 4k, fitting different stages from preview to final cut.

Capability Description

First-frame and first/last-frame driving: input 1 image to generate a video starting from that image, or input 2 images to control the first and last frames; the output aspect ratio follows the first frame, making it easy to match the composition of existing assets.

Subject reference and consistency: define up to 3 named subjects via elements, each with 2–4 reference images and a short description, and mention them in the prompt using @name to help keep subject appearance stable across multiple shots.

Native audio generation: when the audio parameter is enabled, video generation simultaneously produces synchronized sound, suitable for final cuts that need dialogue or ambient sound and reducing audio-video compositing steps.

Intelligent multi-shot planning: when prefer_multi_shots is enabled, the model automatically splits the prompt into multiple shots, obtaining a short film with shot transitions in a single call rather than a single continuous scene.

Use Cases

• E-commerce product images to showcase videos — use the product main image as the first frame to generate a short video with camera movement for product detail pages or feed advertising assets.
• Social media static posters to dynamic clips — input the design draft's first and last frames to produce vertical or horizontal dynamic videos for direct publishing on social platforms.
• Storyboard first/last-frame previews — pair key frames of film or animation storyboards as first and last frames to generate transition shots for confirming pacing and camera movement in pre-production.
• Multi-shot character clips — use elements to lock a character's appearance and enable multi-shot mode to batch-generate short narrative videos with consistent characters for ads or serial content.

Limitations and Selection Advice

A single generation is capped at 15 seconds; longer narratives require segmented generation and stitching. The output aspect ratio is determined by the first frame, so asset proportions need to be unified in advance. Subject references are limited to 3, and each subject requires 2–4 images of the same subject, resulting in relatively high input preparation cost. Native audio and intelligent multi-shot are optional switches, and enabling them makes the prompt description quality more critical. If only text-driven generation is needed, choose the text-to-video route in the same series; if local replacement based on an existing video is required, that falls outside the scope of this image-to-video route.