kling-v2-master is Kuaishou Kling's V2 Master video model, designed for short-shot creation starting from text concepts or a first-frame image, balancing visual quality and user experience. It is suitable for product showcases, concept storyboards, and animated photo assets, and can also handle animation tasks in talking photo workflows. When choosing, focus on single-shot expression rather than multi-asset editing or native audiovisual generation.
Submit requests to the public API at api.acedata.cloud using the documented parameters, then use the results in your application.
Input parameters and result formats vary by service. Use the public API for this model and follow its guide for generation, task retrieval and editing operations.
The durations, aspect ratios, and asset requirements above correspond to this platform's supported usage scope. Talking photos are a combined creation workflow and are not equivalent to the model's native audio accompaniment.
Text-to-video is suitable for turning scene descriptions into watchable shot drafts. When creating, prompts can be organized around the subject, action, environment, and lighting, focusing one piece of content on a clear event. The balanced positioning of V2 Master is suitable for comparing different creative expressions, then using selected results for storyboard discussions or subsequent editing.
Image-to-video uses an existing image as the opening foundation, then describes the desired action through prompts. Compared with starting entirely from text, this approach allows product images, portraits, or scene artwork to directly participate in creation. It emphasizes developing the visuals from a specified starting point and does not provide end-frame locking, making it better suited to short shots whose endings can develop freely.
The talking photo workflow accepts a portrait photo and existing audio, first animates the photo, then applies lip-sync processing according to the audio. After selecting V2 Master, it handles the animation generation portion. This workflow is suitable for character introductions or brief spoken segments with existing voiceovers, without mistakenly writing dialogue as an instruction for the video model to automatically generate sound.
Input a product display image and describe a slight rotation, background atmosphere, or presentation action to generate dynamic clips for short-video editing. Choose landscape, portrait, or square aspect ratios according to the placement, then add selling-point captions and music during editing. It is recommended to highlight one presentation focus at a time rather than arranging an entire advertising storyline within the same shot.
Organize a single scene from the script into descriptions of the subject, action, and environment, then use text-to-video to create shot drafts; when art direction already exists, you can also start from a scene image. The deliverables are short clips that facilitate team discussion, suitable for comparing mood, composition, and action expression before deciding which content enters formal production or manual refinement.
Prepare a clear front-facing photo of one person and a voiceover, then use the talking photo entry point to create character introductions, character greetings, or brief explanations. The audio should stay within the selected video duration, and prompts should mainly describe expressions and actions. After completion, obtain the final video link and integrate it into existing publishing or editing workflows.
V2 Master is suitable for short-shot tasks driven by text or a first frame; V2.1 Master places greater emphasis on quality and consistency. Both use a single mode here, with durations of 5 or 10 seconds, and neither provides an end frame, audio, or structured camera movement. If visual consistency matters, compare results using the same material rather than treating version numbers as a direct guarantee of performance.
If the ending frame must be locked, consider V2.5 Turbo pro, which supports end frames; if synchronized audio is needed, choose V2.6 pro or V3. Multiple images, reference videos, and editing existing videos are better suited to O1 or V3 Omni. The value of choosing V2 Master lies in its clear text-to-video and first-frame creation workflow, not in covering these different control requirements.
Use text2video for text only; use image2video and start_image_url when you already have key visuals. Choose one subject action, and describe the lighting, composition, and appearance that should be preserved.
Specify /kling/videos with model=kling-v2-master, use this model's single mode, and select a duration of 5 or 10 seconds. Image-to-video starts from the first frame, without adding an end frame or a multi-material list.
Use async or callback_url to retrieve task results, and save the task_id and video ID; focus on checking materials, subject deformation, and inter-frame continuity. The output does not include native audio, so voice-over and music should be produced separately.
Use an image of a white ceramic cup as the first frame. The camera slowly pushes in, steam rises from the cup opening, and the background lighting remains stable.
Review based on a short shot of 5 or 10 seconds. Do not provide an end frame or camera_control; if sound is needed, add voice-over separately or use a supported photo talking-head workflow.
Explicitly specify model=kling-v2-master in /kling/videos or /kling/talking-photo. This clearly selects V2 Master rather than relying on the endpoint default. kling-v2-1-master is another model and should not be used as an alternative name for the same model.
Use /kling/videos, set action to image2video, provide start_image_url, and use prompt to describe the action and scene changes. Choose a duration of 5 or 10 seconds. This model does not support end-frame control, so the first frame should be treated as the starting point rather than constraining both the start and end points.
You cannot use a video prompt as speech generation input. To create a talking photo, provide image_url and audio_url to /kling/talking-photo; prompt is used for actions or expressions during the photo animation stage. It is recommended to prepare the voiceover first, then have the photo perform lip-sync processing based on that audio.
This model generates videos with durations of 5 or 10 seconds, and talking photos also offer these two duration options. Longer content can be split into shots and then edited together; if you need more flexible single-generation durations, consider V3 instead of passing arbitrary seconds to V2 Master.
You can use async=true to obtain task_id and then query the task result; you can also set callback_url to receive completion notifications. The application should associate the task ID with the creation record and read state and video_url upon completion. Successful task submission and completed video generation are different stages and should not be treated as the same status.
Choose a development approach or application suited to this model and follow its setup guide.
Python SDK Integration Guide
SDK installation, authentication and examplesTypeScript SDK Integration Guide
SDK installation, authentication and examples
