Kling API

Kling 3.0 Turbo is a video generation model from Kuaishou's Kling team, released as a faster, lower-cost variant in the Kling 3.0 family. It generates video with synchronized audio and is tuned for quick turnaround in high-volume workloads. The release notes highlight improved lip-sync, which makes it a fit for talking-head and dialogue clips. Through the API it supports 720p and 1080p output at 24 FPS, with clip lengths from 3 to 15 seconds. Choose it when generation speed and cost per clip matter more than maximum fidelity. Third-party comparisons say the original Kling 3.0 remains the better pick for 4K output and its storyboard tooling. It suits developers building apps that generate many short clips, such as marketing videos, social content, and spoken-word scenes.

Kling Image 3.0 is a text-to-image model from Kuaishou's Kling AI, released as part of the Kling 3.0 series alongside Video 3.0, Video 3.0 Omni, and Image 3.0 Omni. Kuaishou describes the series as having major upgrades in consistency and photorealistic output. It did not publish benchmark numbers for the image model in the sources we found, so we do not cite any here. Through the API it supports eight aspect ratios (16:9, 9:16, 1:1, 4:3, 3:4, 3:2, 2:3, and 21:9), which covers landscape, portrait, square, and ultrawide layouts. It costs $0.028 per 1k image. It suits developers who want photorealistic output from a single model at a low per-image price, and who need flexible output shapes for web, social, or banner use.

Kling 3.0 is a video generation model from Kuaishou, released in February 2026 as the successor to Kling 2.6. It generates video from text prompts and reference images, and produces synchronized audio together with the video. It supports multi-shot generation, so a single prompt can produce a clip with several cuts, and clips run up to 15 seconds. An Elements feature lets you supply reference images or clips to keep characters consistent across shots. Audio includes dialogue bound to specific characters in English, Chinese, Japanese, Korean and Spanish. On Puter, it accepts durations from 3 to 15 seconds and outputs at 720p, 1080p or 4k. Kling 3.0 suits developers building short narrative clips, character-driven scenes or ads where multiple shots and spoken audio need to come from one API call.

Kling 3.0 Omni is a multimodal video generation model from Kuaishou's Kling AI, released as part of the Kling 3.0 series. It accepts text, images, audio, and video as input and combines text-to-video, image-to-video, reference-to-video, and in-video editing in a single model. It generates clips of up to 15 seconds with native audio, including speech in multiple languages, dialects, and accents. Kuaishou describes the 3.0 series as improving consistency across shots, and the Omni variant is the one that takes reference inputs. Through the API, durations from 3 to 15 seconds are available at 720p, 1080p, and 4K output. Choose Kling 3.0 Omni when a project needs reference-guided generation or editing of existing footage in addition to plain prompts. It suits developers building video tools that need longer clips with synchronized audio from one endpoint.

Kling 2.6 is a video generation model from Kuaishou's Kling AI. It generates video and matching audio in a single pass, including voiceovers, sound effects, and ambient sound, so no separate dubbing step is needed. It supports text-to-video and image-to-video with audio. Kuaishou lists Chinese and English voice generation, and clips run up to 10 seconds. Through the API, durations are 5 or 10 seconds at 720p or 1080p and 24 fps. It is a fit for developers who want short clips with synchronized speech and sound from one API call, such as ads, social content, and product demos. It is the first Kling release with native audio, and Kuaishou describes it as replacing the older workflow of silent video followed by manual audio.

Kling O1 is a unified multimodal video model from Kuaishou's Kling AI, released in December 2025. It takes text, images, video, and subject references as input and handles generation and editing in one model. It supports text-to-video, reference-based generation, start and end frame generation, video inpainting, style re-rendering, and shot extension. Edits can be requested in a single prompt, such as swapping a character, changing the weather, or adjusting colors, without masking or manual keyframing. Clips run from 3 to 10 seconds at 720p or 1080p and 24 FPS. Kuaishou describes it as built on a Multimodal Visual Language framework. Developers who want generation, reference-driven shots, and video editing behind one API model can use Kling O1 instead of chaining separate models.

Kling 2.5 Turbo is a text-to-video and image-to-video generation model from Kuaishou's Kling AI. It supports 5 and 10 second clips at 720p and 1080p, 24 FPS. Kuaishou says it improves on earlier Kling versions in prompt adherence for multi-step instructions and scene transitions, motion and camera movement, and physics in fast action. In image-to-video, it aims to preserve the art style, colors, and lighting of the source image. In Kuaishou's own blind test with professional voters (not independently verified), the reported win-loss ratios for text-to-video were 285% against Seedance 1.0 mini, 212% against Veo3-fast, and 160% against Seedance 1.0. It suits developers who want short, controllable clips from a prompt or a still image through one API.

Kling Image 2.1 is a text-to-image generation model from Kuaishou, the company behind the Kling AI video and image models. It is described as a high-fidelity model known for detailed photorealistic output, natural lighting, and strong character consistency. The model supports eight aspect ratios (16:9, 9:16, 1:1, 4:3, 3:4, 3:2, 2:3, and 21:9), which covers landscape, portrait, square, and ultrawide formats without cropping afterward. Through Puter it costs about $0.014 per 1k image, so it is inexpensive to use for batches and iteration. It suits developers who need photorealistic images or consistent characters across generations, such as portraits, product and food scenes, interiors, and marketing visuals. No official benchmark scores have been published for this model.

Kling Image 3.0 Omni is an image generation model from Kuaishou's Kling AI, released as part of the Kling 3.0 series alongside Video 3.0, Video 3.0 Omni and Image 3.0. Kuaishou describes the 3.0 series as a major upgrade in consistency and photorealistic output, with multimodal input across text, images, audio and video. It has not published image-specific benchmark numbers or direct comparisons with other image models. Through the API it supports aspect ratios of 16:9, 9:16, 1:1, 4:3, 3:4, 3:2, 2:3 and 21:9, at about $0.028 per 1K image. It suits developers who want text-to-image output with a wide range of ratios, including ultrawide, in the same model family as Kling's video models.

Kling Image O1 is a multimodal image generation and editing model from Kuaishou, the company behind Kling AI. It was released alongside the Kling O1 video model. The model accepts a text prompt alone or a prompt with up to 10 reference images. It is built for editing existing images (adding, removing, or changing details), controlling style, and keeping subjects and visual features consistent across outputs, such as character sets or comic panels. Through Puter it supports aspect ratios of 16:9, 9:16, 1:1, 4:3, 3:4, 3:2, 2:3, and 21:9, at $0.028 per 1k image. Kuaishou has not published benchmark numbers that we could find. It suits developers who need consistent characters or products across many generated images, or who want to edit images from a text instruction through one API.