Two ways to clone a voice:
Learn more about Voice Clone + TTSInstant clone is one-shot — nothing is stored. Permanent clones live in the API tab's flow guide.
Your cloned voice shows up here
Record a minute or two, type a line, hit the button.
Two ways to clone a voice:
core-voice-clone, asynchronous): upload an audio sample, submit a job, and get back a reusable voice_id you can pass to core-tts whenever you like. One 1500 credits flat charge per clone; the TTS calls you make afterwards are billed at standard per-second rates.core-instant-clone-tts, synchronous, 200 credits/request): one shot — send an audio URL plus the text and get the speech straight back. By default the body is a binary audio/wav stream; pass stream: false to have it uploaded and get a JSON envelope with a url instead. No reusable voice is stored.Audio samples are passed as URLs. Use CN8's upload-media to upload a local file and get a public URL. If the source is a video, the audio track is extracted automatically (up to ~60 seconds).
Submit a clone job to core-voice-clone and get a 202 with a job_id. When the job completes, the result includes a reusable voice_id (plus a message and the provider). Reuse the voice_id with core-tts at standard pricing.
core-instant-clone-tts clones the voice and synthesizes in one request. The response is binary audio/wav by default, with metadata in custom headers (X-Used-Voice-Id, X-Provider, X-Text-Length, X-Audio-Duration); send stream: false to get a JSON envelope with a hosted url instead.
audioUrl can point to either an audio file (mp3/wav/ogg) or a video file — for video, up to 60s of audio is extracted before cloning.
Defaults to elevenlabs; minimax also supports cloning. The cloned voice_id is tied to the provider that created it — pass that same provider when you use it later in core-tts.
Brand voice
Record a spokesperson once, clone, then generate all future ads / promos / announcements via core-tts in that voice.
Character voices
Clone game / audiobook character voices and generate dialogue at scale.
Quick one-off clip
Use instant clone when you need a single audio output without storing a voice profile.
Localization with same speaker
Clone the original speaker once, then synthesise dubbed lines in multiple languages keeping voice character (combine with TTS or Dubbing).
input
audioUrl (audio or video) + provider; for instant clone also text, optional voice_settings, and optional stream (true = audio bytes, false = hosted url)
output
Persistent clone: a 202 with a job_id (poll /v2/jobs/{job_id} for result_data.voice_id). Instant clone: a binary audio/wav stream plus metadata headers by default, or a JSON envelope with a hosted url when stream is false.
core-voice-clone is asynchronous: you get a 202 with a job_id right away. Poll GET /v2/jobs/{job_id} until status = completed (typically 1-3 min). result_data carries voice_id, message, and provider. Save the voice_id and pass it as voice_id to core-tts in your later calls.
audioUrl can point to a video file. The audio track is extracted automatically (about 60s) and used as the clone source. Use upload-media with media_type=video to upload a video and get a URL.
audioUrl is camelCase (not audio_url). Both core-voice-clone and core-instant-clone-tts use this spelling. Sending audio_url returns a 422.voice_settings is snake_case (matching core-tts).core-instant-clone-tts returns binary audio/wav by default — the Content-Type and X-* headers carry all the metadata. Send stream: false if you would rather receive a JSON envelope with a hosted url, the same shape core-tts returns.audioUrlrequiredstringURL to a short voice sample (camelCase).
textrequiredstringText to synthesise with the cloned voice (max ~5000 characters).
providerstring · default: elevenlabsSynthesis provider
enum: elevenlabs, minimax
voice_settingsobjectProvider-specific voice settings — same shape as core-tts.voice_settings (stability, similarity_boost, etc.).
streamboolean · default: trueHow you want the audio back. true (default) streams raw audio/wav bytes with metadata in headers — best for a browser that plays the result immediately. false uploads the audio to storage and returns the same JSON envelope as core-tts, with a url you can store or hand to another service — best for server-side integrations, which would otherwise have to buffer and re-host the blob themselves.
(body)binarystream=true only — raw .wav audio bytes (Content-Type: audio/wav).
urlstringstream=false only — the hosted audio. A presigned link when the bucket is private, otherwise a direct URL.
keystringstream=false only — the storage key of the uploaded audio.
duration_secondsnumberstream=false only — length of the generated audio. This is the billed quantity.
voice_idstringstream=false only — the voice that was cloned for this call (same value as the X-Used-Voice-Id header).
X-Used-Voice-Id (header)stringThe voice ID that was generated for this call.
X-Audio-Duration (header)stringstream=true only — length of the generated audio in seconds. This is the billed quantity; a binary body has no field to carry it.
X-Provider (header)stringThe provider that produced the audio.
X-Text-Length (header)stringLength of the input text (in characters).
X-Gateway-Token-Cost (header)stringAmount deducted from your balance for this call.
X-Gateway-Credit-Balance (header)stringYour remaining balance after this call.
Persistent clone is a one-time per-clone fee; instant clone and all TTS synthesis are billed per second of generated audio at the standard rate.
| Service | Unit | Price |
|---|---|---|
| Persistent Clone | item | 1500 credits/clone (one-time) |
| Instant Clone + TTS | second | 4 credits/second |
| TTS with cloned voice | second | 4 credits/second (standard core-tts rate) |
| List Voices | item | Free |
Synthesis costs the same either way (4 credits/s); the difference is the one-time 1500 credits persistent-clone fee. Instant clone is cheaper until cumulative usage justifies storing the voice; persistent pays off when you'll reuse the same voice many times (reusable voice_id, no clone step per request).
Instant clone returns BINARY audio/wav — there is no audio_url to download separately; the response body itself is the audio.
Voice cloning is supported on ElevenLabs and MiniMax. OpenAI does not support cloning (supports_voice_cloning=false in the catalog).