Skip to main content
POST
Create a private voice
Give the voice a display name and set sampleUrl to a public HTTPS link to your own recording. Use 5–30 seconds of clean speech from one speaker whose voice you have permission to replicate. MP3, WAV, MP4 and MOV are supported.
  • HTTP 202, pending — Copy operationId into Get voice operation to check progress.
  • HTTP 200, active — Use voiceId in assets.voices[].voiceId and listen to previewUrl if provided. See the Omni request example.
  • HTTP 200, failed — Read message for the reason. Submit a corrected sample as a new creation.
Billing and limits. Sending a new request in Try it can charge your wallet. Each accepted creation is billed once at your applicable Custom voice creation rate, including failed clones and unresolved upstream responses. Kling’s official base rate is $0.007 per request. Rejections before acceptance are free. Listing, status checks and deletion are free. Referencing the voice in Omni adds no separate voice fee. Pricing Retrying a request. The Idempotency-Key header is optional. Without it, each accepted submission creates a separate operation and incurs a charge, even for the same sample. For safe retries, send a key on the first request and reuse it with the same body. GENGEN then returns the original operation without another charge. Reusing a key with different input returns HTTP 409. Each workspace has five private voice slots. Pending and active voices, as well as voices awaiting deletion, occupy slots.

Authorizations

Authorization
string
header
required

GENGEN API key

Headers

Idempotency-Key
string

Optional. To retry safely, choose a key before the first request and reuse it with the same body in this workspace. Use printable ASCII without spaces.

Required string length: 1 - 128
Pattern: ^[!-~]+$

Body

application/json
name
string
required

A display name for the private voice, measured after trimming surrounding whitespace.

Required string length: 1 - 20
Example:

"Narrator"

sampleUrl
string<uri>
required

Public HTTPS URL for 5–30 seconds of clean speech from one authorized speaker. Supply an MP3, WAV, MP4 or MOV file.

Maximum string length: 4096
Pattern: ^https://
Example:

"https://example.com/speaker.wav"

Response

Current state of a completed creation or an idempotent replay. Inspect status before using voiceId.

operationId
string<uuid>
required

Stable workspace voice operation ID. Use it in the voice-operations path to check progress.

status
enum<string>
required

Pending and deleting require polling. Active means ready for Omni, failed means creation failed, and deleted confirms removal.

Available options:
pending,
active,
failed,
deleting,
deleted
name
string
required

Display name supplied at creation.

voiceId
string

Returned when active. Copy this string into an Omni request or the delete endpoint.

Required string length: 1 - 128
Pattern: ^[A-Za-z0-9_-]+$
previewUrl
string<uri>

Optional HTTPS link to an audio preview. Open it to listen to the voice.

message
string

Additional detail about a failure or a deletion awaiting confirmation.