What does a pro clone cost?

Training a pro clone costs 1M credits as a one-off. Generation then runs at 1.2 credits per character, against 1 credit per character for instant clones and catalog voices. Instant clones carry no training cost. Clone limits per account scale with plan tier.

Can it reproduce a specific regional accent?

Yes. Regional accents clone faithfully, including Munich, Frankfurt and Berlin German, Québécois French and Scottish English, because the clone reproduces what is in the sample rather than mapping it onto a standard accent. The corollary is that whatever the sample contains, you get: a rushed or noisy recording clones as a rushed, noisy voice.

Do we need the speaker's permission?

Yes, without exception. Instant cloning is gated behind sign-in for that reason, and pro clones are built with the voice owner involved. If voice data cannot leave your infrastructure, the in-VPC deployment keeps audio inside your own AWS account. Full policy at http://trust.gradium.ai.

Spotlight post

Gradium Extends Funding to $100 Million and Expands to Silicon Valley

Gradium extends its seed funding to $100 million, welcoming new investors including NVIDIA, and opens a San Francisco Bay Area office to scale its real-time voice AI.

July 8, 20263 min read

Gradium Extends Funding to $100 Million and Expands to Silicon Valley

Models

Text-to-SpeechTurn text into natural sounding voice Speech-to-TextTranscribe audio into text instantly Live TranslationConvert voice to voice in real time Voice CloningInstance or pro voice clone Voice DesignA brand-new voice from a text prompt On-Device TTSOffline real-time natural voice

Showcase

Agent DemoTest our voice agent live GradbotBuild voice agents in a single prompt CommunityBuilders, researchers and creators App GalleryApps built on Gradium

Clone any voicefrom 10s of audio

Top rated instant clones that keep the rasp, breath and pitch dynamics of the original.

Moritz

Original voice

Cloned voice

Clone a voice free Request a pro voice clone

The numbers that make voice agents feel real

4 / 4
Languages won
Highest Elo vs ElevenLabs Flash 2.5 in English, French, Spanish and German, in 3,220 blind human comparisons.
10 s
Of audio to clone a voice
No training step. Your voice ID is ready in seconds.
1,000
Instant clones a month
On every paid plan, from $13/month. Unlimited on Enterprise.
1,000Instant clones a month

Try now Talk to us

No credit card required

Why clones sound fake elsewhere

Most providers bolt cloning on as a layer through prefixing, which needs transcripts and clean audio. Gradium reads the recording directly, so clones keep vocal fry, rasp and pitch dynamics in any language.

Why your voice cloning sounds fake

Blind A/B evaluations

3,000+

Languages won on Elo

4 of 4

Voices × sentences

20 × 900

No marketing spin,just code and hard numbers

Paste one of these prompts into your coding agent and it will read the guide, install the SDK and open the stream. Nothing else to do, you're ready to go.

1Add Gradium instant voice cloning to this project.
2
3Read https://docs.gradium.ai/api-reference/introduction first and follow it
4over anything you assume. Use the official Gradium SDK for this codebase's
5language if it wraps the voices endpoint; otherwise call the REST endpoint
6directly. Read the API key from an environment variable, never hardcode it.
7
8Then implement the flow, in this order:
9
101. Take an audio sample and a name, and POST them as multipart form data to
11   the voices endpoint with the language of the sample. Reject samples shorter
12   than 10 seconds before uploading rather than after.
132. Use start_s to skip any lead-in when the caller passes a longer recording,
14   so the clone is built from speech rather than room tone.
153. Read uid from the response and store it. That value is the voice_id for
16   every other endpoint; do not keep a second identifier for the same voice.
174. Synthesise with that voice_id through the standard TTS endpoint, exactly as
18   you would with a catalog voice.
19
20Keep it small: one module, typed, with clear errors when the sample is too
21short, when the language is outside the supported five, or when the key is
22missing. Do not add a UI unless I ask for one.

No distortion under load

Latency stays flat as concurrency climbs, flexible deployment that fits your setup.

Real-time API

Bidirectional WebSocket streaming. Start sending before ready returns; first audio in ~200 ms. Telephony formats built in.

in · pcm 24 kHz

CLIENT/AGENT

WEBSOCKET

Gradium API

wss://api.gradium.ai/
api/speech/s2s

BidirectionalTelephony ulaw / alaw~200 ms

out · pcm 48 kHz

VOICE ENGINE

Copied!

Predictable & secure

Flat latency under load, enterprise SLAs, zero data retention, ISO 27001 certified.

1 conversationQuiet hour

I’d like to move my appointment.

50ms

Of course. Which day works?

10 000 conversationsPeak

I’d like to move my appointment.

50ms

Of course. Which day works?

Runs where you need it

Cloud API, dedicated instances, self-hosted, on-prem for data sovereignty, and cloud-provider marketplaces.