Kling Subjects and Voice Assets

POST /kling/elements and POST /kling/voices manage platform assets through action: list lists your own assets, presets obtains preset assets, retrieve queries an id, and delete deletes your own id. Preset assets cannot be deleted. All references use the platform IDs returned by the API; assets are isolated by user and application. Management operations are not billed.

list and presets support page_num (1–1000, default 1) and page_size (1–500, default 30). Preset assets can be used for generation after being read for the first time. Deleted assets can no longer be used for new tasks.

¶ Create Voice

Submit action="create", a voice_name of up to 20 characters, and voice_url to /kling/voices. Audio should be a clear single-person voice of 5–30 seconds, supporting mp3, wav, mp4, and mov. Each creation costs 0.07 Credits; failed tasks follow the platform's failed-task billing rules. When async=true or callback_url is provided, task_id is returned first, and the final result is queried through /kling/tasks.

{"action":"create","voice_name":"Narrator","voice_url":"https://cdn.acedata.cloud/assets/examples/fish/5ade0339-5f11-487e-aacc-06a908271706-8e3fcb0e5547.mp3","async":true}

After successful creation, use voice_list=[{"voice_id":"platform voice ID"}] in /kling/videos. It supports 1–2 voices and is limited to model="kling-v2-6", mode="pro", and generate_audio=true. Use <<<voice_1>>> or <<<voice_2>>> in prompt to specify the selected voice. Videos with specified voices are billed at 1.68 Credits/second based on actual output duration, supporting fractional seconds; usage.duration in the result is the actual metered duration. For example, an output of 5.04 seconds corresponds to 8.4672 Credits. It cannot be mixed with element_list. A 502 is returned when the service does not provide a valid output duration.

The estimated total price before generation is calculated based on the requested duration: 5 seconds costs 8.4 Credits, and 10 seconds costs 16.8 Credits; the final bill is settled based on the actual output duration.

¶ Subject References

Use element_list=[{"element_id":"platform subject ID"}] in /kling/videos to reference obtained preset subjects. It supports 1–3 subjects and is limited to V3, V3 Omni, and O1. For Omni/O1, the total number of subjects plus images is up to 7; when there is a reference video, it is up to 4. Other generation limits still apply.

Custom subject creation is not yet available; action="create" returns 503; creation tasks will not be submitted before the price is confirmed. The request contract under preparation includes element_name (up to 20 characters), element_description (up to 100 characters), and reference_type. image_refer uses a front-facing image and 1–3 reference images; video_refer uses one reference video of 3–8 seconds, 1080p, and up to 200 MB. The two material forms cannot be mixed. This preparation contract does not indicate that the creation capability has been launched.