In sintesi (BLUF): Kling 4.0 Flash is a lighter and more affordable version of Kling 4.0: clips of 3–20 seconds, 720p output, the same set of input materials, but faster and cheaper. Great for trying variants, scene drafts and bulk generation where speed matters.
The model accepts a text prompt, photos, video references and elements — up to 15 references in one request (up to 10 images and 5 videos), just like the base Kling 4.0. Camera movement and scene pacing are carried over from the video reference.
Key differences from Kling 4.0: duration 3–20 seconds, 720p output only, keyframe mode supports a single first frame, voices and keep-original-sound are not supported. In exchange — a notably lower price and faster generation.
Duration 3–20 seconds, aspect ratios auto, 16:9, 9:16, 1:1 and 21:9. The price is computed transparently before the run and stays frozen: with video references both the output seconds and the input video seconds are billed (whichever is larger).
Tokens are charged when the task is created and refunded automatically if the generation fails.
The Video section on the site — pick Kling 4.0 Flash in the model list. In the public API — POST /v1/videos/generations with the kling-4-0-flash model (reference mode, or keyframe mode with a single first frame); media is passed via links or Base64, and the finished video is polled by task id.
Own applications can use the model through the public API or MCP (the generate_video tool). Get a key on the API page.
Flash is the fast, cheaper version: 3–20 second clips, 720p only, a single first keyframe, no voices and no keep-original-sound. References (up to 15) work the same way.
Billed in tokens; the exact price is shown before the run and stays frozen. With an input video both the output and input seconds are billed (the larger one).
Voice generation and keeping the original audio track are not supported in Flash; for native sound and lip-sync use the base Kling 4.0.
Flash outputs 720p; for 1080p use the base Kling 4.0.