Kling Video 4.0 Flash Reference-to-Video.

Kling Video 4.0 Flash Reference-to-Video Overview

Kling Video 4.0 Flash Reference-to-Video generates video from a text prompt with optional image, video and Element references. This model is part of the Kling Video 4.0 Flash family. Structured reference roles distinguish source assets instead of treating every upload as an opening image. Supply image references, feature or base videos, and existing Elements, with adaptive or explicitly selected framing.

Start with a required prompt and add the optional image, video or Element references needed for the scene. Choose duration, resolution and aspect_ratio. Reference arrays are optional, but their combined limits apply whenever supplied. Output durations run from 3 to 20 seconds at 720p. Use the model for short creative clips, concept development and application workflows where the intended action fits a single short scene. Review the result for the details that matter to your project before selecting it as a finished asset.

Capabilities

• Guide video generation with a required prompt and optional structured references.
• Supply up to ten images with first_frame or refer_image roles.
• Supply up to five videos, including at most one base_video.
• Reference up to seven existing Elements, including at most three Video Character Elements.
• Select a 3–20 second output duration at 720p.
• Choose adaptive, 21:9, 16:9, 9:16 or 1:1 output framing.
• Distinguish supplied assets with unique reference identifiers.

Use Cases for Kling Video 4.0 Flash Reference-to-Video

Creators can build a scene around a known visual subject. "The referenced character crosses the courtyard while the camera follows at walking speed" explains how the subject should participate instead of leaving the reference role implicit.

Marketers can provide separate product and setting images for a concept video. "Place the referenced product on the table in the setting reference, warm window light, slow camera push" separates the intended contributions. Inspect appearance before preparing final campaign assets.

Designers can combine a first-frame image with supporting references to communicate composition and appearance. Keep assets visually compatible and describe one coherent action instead of asking every reference to supply a different scene.

Developers can build reference-driven video workflows with the Kling Video 4.0 Flash Flash Reference-to-Video API, including existing Element IDs when an application already has those assets.

Tips and Tricks

Give each referenced asset a short, unique id and explain its role in the scene. Separate subject, setting and motion instructions so different references do not compete to define incompatible details.

For a product reference, try: "Show the referenced bottle on a sunlit desk. Keep its overall appearance recognizable while the camera slowly moves closer." For a character reference: "The referenced character walks through a quiet courtyard at sunrise. Gentle forward tracking shot, one continuous action."

Add only references that contribute useful information. Multiple refer_image entries require ids, and all supplied ids must be unique across images, videos and Elements. Keep the combined count at 15 or fewer. Start with one reference, inspect the result and add further guidance only when it addresses a specific creative need.

Technical Specifications

• Output: generated video; a successful task provides a video URL.
• Duration: integer seconds from 3 through 20; default 10 seconds.
• Resolution: 720p only; default 720p.
• Required input: prompt. Optional references: images, videos and elements.
• Images: up to ten JPG/JPEG/PNG files, 30 MB each, minimum 300 × 300 pixels; aspect ratio 2:5–5:2.
• Videos: MP4/MOV, up to 300 MB each, 3–20 seconds, 24–60 fps; dimensions 700–4553 pixels, maximum 8,294,400 pixels.
• Framing: adaptive by default; 21:9, 16:9, 9:16 or 1:1.
• Prompt limits: English-only 1–26,000 characters; non-English-only up to 11,000. Mixed-language input must satisfy E/26000 + N/11000 ≤ 1; Kling enforces the weighted limit.
• Asynchronous. No fixed completion time, output codec or configurable output frame rate is specified by the current integration.

Things to Be Aware Of

The array maximums cannot all be filled at once: images, videos and Elements together must not exceed 15. Videos plus Video Character Elements must not exceed five. Combined reference-video duration, including associated character clips, must stay within 3–20 seconds. Kling checks actual media properties and Element-dependent rules at runtime, so valid JSON does not ensure every supplied asset will be accepted.

Respect the language-weighted prompt limit and send only documented fields. Describe unwanted details in the same prompt instead of adding a negative_prompt parameter. Review a simple request before preparing variations.

Key Considerations

Give each reference a clear role and a unique id. The default framing is adaptive, with four explicit aspect ratios also available. Check total counts and associated video durations before submission. Begin with a small reference set so the influence of each asset remains understandable.

Billing is calculated per task from the reported charge, rather than a published flat price per clip. No fixed per-second estimate is present in this model’s current catalog. Review the final billed cost before budgeting a large batch.

Limitations

References guide generation without guaranteeing exact identity, product geometry or pixel-level reproduction. This endpoint accepts existing Element IDs but does not create Elements. It accepts no voice references, middle-frame or last-frame roles. At most one base_video is supported. Output is limited to 720p and 3–20 seconds, with no seed or separate output frame-rate control. Check important details before publication.