ElevenLabs Voice Cloning API: Endpoints, Steps & Costs (2026)

Yes. POST /v1/voices/add creates an Instant clone from uploaded audio, and POST /v1/voices/pvc starts a Professional clone with captcha verification and training. API cloning follows your plan: Starter+ for Instant, Creator+ for Professional. Generation costs 1 credit per character. Clone only your own voice or one with written consent. Checked 4 Nov 2026.

Can you clone voices through the ElevenLabs API?

Yes. ElevenLabs’ published OpenAPI specification includes dedicated voice cloning endpoints for both Instant and Professional cloning. Every call authenticates with an xi-api-key header, and feature access follows your plan — the API does not bypass plan limits.

Two workflows exist. Instant cloning is a single call: POST /v1/voices/add with your audio files as multipart/form-data, and the voice is ready in seconds. Professional cloning is a multi-step pipeline: create the voice record, upload samples, pass captcha verification, then start training.

If you are new to the API generally, our API quickstart guide covers authentication and first calls; this page focuses on the cloning endpoints.

One practical note: ElevenLabs’ dashboard documentation requires sign-in, but the OpenAPI specification at api.elevenlabs.io/openapi.json is publicly accessible without an account. That specification is the source for every endpoint name and purpose in this guide, and it was checked on 4 Nov 2026 — use it as your reference when the dashboard docs are out of reach.

Which endpoints handle voice cloning?

Endpoint names and purposes come from ElevenLabs’ published OpenAPI specification at api.elevenlabs.io/openapi.json, checked 4 Nov 2026:

How to clone a voice via API, step by step

Instant cloning first — the path most developers take:

1. Get an API key from your ElevenLabs account and include it as the xi-api-key header on every call. Cost: 0 credits.
2. Prepare 1–5 minutes of clean, single-speaker audio. Cost: 0 credits.
3. POST /v1/voices/add with the audio as multipart/form-data plus a name for the voice. The clone is ready in seconds. Cost: 0 credits to create.
4. Generate a 30-second test of about 450 characters with your usual text-to-speech call to verify the match. Cost: about 450 credits.
5. Generate production audio at 1 credit per character — a 10-minute video of about 8,500 characters costs about 8,500 credits.

Professional cloning via API follows the longer pipeline:

1. POST /v1/voices/pvc to create the voice record with its metadata. Cost: 0 credits.
2. POST /v1/voices/pvc/{voice_id}/samples to upload 30+ minutes of clean audio as multipart/form-data. Cost: 0 credits.
3. GET the captcha, solve it, and POST it back to /v1/voices/pvc/{voice_id}/captcha — or POST to /verification for manual review. Cost: 0 credits.
4. POST /v1/voices/pvc/{voice_id}/train to start training. ElevenLabs publishes no exact turnaround and no separate training credit cost.
5. Generate with the trained voice at 1 credit per character.

How does API verification work for Professional clones?

The OpenAPI spec exposes the verification machinery directly: a captcha retrieval endpoint, a captcha submission endpoint (multipart), and a manual verification request endpoint. This is the API equivalent of the dashboard’s enterprise-grade voice verification.

Verification exists to enforce consent — the API confirms you are authorized to clone the voice before training starts. Captchas are designed to resist automation, so build the human step into your pipeline rather than scripting around it.

When verification fails, training does not start. The usual causes mirror the dashboard: samples that are too short, noisy, or multi-speaker, or a voice that does not match the authorized speaker. The manual verification endpoint exists for edge cases — legitimate voices that automated checks cannot confirm — so a failed captcha is a pause, not a ban.

What does API cloning cost in credits?

Creating the clone costs nothing; generating audio costs 1 credit per character on standard models. Uploads, verification, and training carry no published credit cost.

API users get discounted rates on some models: V2 Flash/Turbo English and V2.5 Flash/Turbo Multilingual cost 0.5–1 credit per character via API, per the official pricing FAQ. High-volume pipelines should prefer these models.

Budget example: cloning via API on Starter ($6/month, 30,000 credits) supports about 30 minutes of generated audio monthly; on Creator ($22/month, 121,000 credits) about 2 hours.

Which plans include API cloning?

API access follows plan features: Instant cloning via API needs Starter ($6/month) or higher; Professional cloning via API needs Creator ($22/month) or higher with its 1 PVC slot. The free plan cannot clone through the API or the dashboard.

Concurrent request limits (3 on Starter through 25 on Business) matter for API pipelines generating at scale — check your plan’s cap before building.

Higher plans also raise the practical ceilings around API cloning: more custom voice slots (160 on Pro, 660 on Scale, 2,200 on Business), more professional slots (3 on Scale, 10 on Business), and larger credit pools for generation. A pipeline that creates a new professional voice per client needs Scale or Business for the slots alone.

The dashboard remains an alternative: our step-by-step cloning guide walks through the non-API path.

Consent rules for API cloning

The API does not relax consent: clone only your own voice or a voice with written consent. The verification endpoints exist precisely to confirm authorization before training starts.

If your app clones end-user voices, collect written consent in your own signup flow before calling /v1/voices/add — the API key holder is responsible for what the key creates. Fraud, impersonation, and misleading content are illegal in most jurisdictions.

Who should NOT clone via the API

Anyone without a paid plan, anyone cloning without written consent, and anyone who only needs one voice once — the dashboard is faster for one-offs. API cloning pays off for repeated or user-driven workflows.

Developers should also skip API cloning when the use case is a single voice for a single project — the dashboard flow in our step-by-step cloning guide finishes the job in minutes with no integration work. Reserve the API for workflows that repeat: user onboarding, content pipelines, or multi-voice production.

Frequently asked questions

What is the API endpoint for Instant Voice Cloning?

POST /v1/voices/add. It accepts your audio files as multipart/form-data along with the voice name, authenticates via the xi-api-key header, and returns a working clone in seconds. The endpoint is named ‘Add Voice’ in ElevenLabs’ published OpenAPI specification.

How do I create a Professional voice clone via API?

Follow the pipeline: POST /v1/voices/pvc to create the voice record, POST /v1/voices/pvc/{voice_id}/samples to upload 30+ minutes of audio, complete captcha verification (GET then POST /captcha), and POST /v1/voices/pvc/{voice_id}/train to start training. Manual review is available via POST /verification.

Does API cloning cost extra credits?

No. Creating clones, uploading samples, verification, and training carry no published credit cost. You pay the standard generation rate — 1 credit per character, or 0.5–1 on discounted API models — when you generate audio with the clone.

Can I use the free plan for API voice cloning?

No. API cloning follows the same plan gates as the dashboard: Starter ($6/month) or higher for Instant, Creator ($22/month) or higher for Professional. The free plan’s 10,000 credits cover standard text-to-speech only.

How do I authenticate API cloning calls?

With the xi-api-key header on every request, using a key from your ElevenLabs account. Upload endpoints (/v1/voices/add, /samples, /captcha) use multipart/form-data; metadata endpoints use JSON.

Where is the official API documentation?

ElevenLabs publishes its OpenAPI specification at api.elevenlabs.io/openapi.json, which names every endpoint in this guide. The in-dashboard docs require sign-in, but the specification itself is publicly accessible and was checked 4 Nov 2026.

Sources checked

Endpoint names, purposes, and authentication come from ElevenLabs’ published OpenAPI specification, checked 4 Nov 2026. Plan prices, credit allowances, and per-plan cloning access come from the official ElevenLabs pricing page, checked 4 Nov 2026. Cloning requirements come from the official ElevenLabs voice cloning page, checked 4 Nov 2026.