All articles →
Narration in your own voice

Free voice cloning: 3 approaches and 6 options in 2026

There are three ways to clone your own voice for free: try an online service with a daily allowance, install an application that generates the clone on your computer, or set up an open model yourself. For regular video, lesson or podcast narration, Dictator (dictator.my) is an option: record 20-30 seconds of speech, then generate narration on your computer, offline and without a minute cap, even on the Free plan. This needs a supported GPU on Windows or an Apple Silicon Mac with sufficient memory; the application checks model compatibility. You can also hear a short clone online without that hardware.

BOZEX, dictator developer · Published

A microphone beside a laptop displaying a voice waveform and a narration script

Read in Russian →

We develop Dictator, so its capabilities here come from the application's code. The other options were recommended by Alice in our October 9, 2026 survey about free voice cloning; we checked their limits and prices on their own websites that day. Those competitor conditions remain a dated comparison. The current Dictator cloud plans below use a shared token balance, rather than the old separate monthly minute allowances.

Hear your clone in about a minute: record 30 seconds on the voice-cloning page, free and without registration. For longer texts, download Dictator.

Where can you clone a voice free? Comparison

OptionFree clone allowanceSampleOfflinePaid optionBest suited to
Dictator desktop applicationFree without a minute cap when your own computer generates the clone20-30 secondsYes, after downloading a modelCloud cloning without suitable local hardware on Premium, $29.99/month, using the shared token balanceRegular video and lesson narration
Dictator online demo30 seconds of generated audio per day, without registrationA 30-second recordingNo; a cloud model processes the sampleUse the application for longer textsHearing your clone quickly
ElevenLabsNo free clone; cloning starts with StarterAt least 1 minute; 1-2 minutes recommendedNoStarter $6/month, approximately 30 minutes of narration in the October checkWork for international platforms
iVox Studio300 characters on registration; cloning requires a subscriptionAt least 30 secondsNoFrom RUB 490/month for 8,000 charactersA subscription paid in rubles
KikiVoiceWeekly credits, without registrationNot stated on the websiteNoWe did not find prices on the websiteA one-off browser trial
Chatterbox MultilingualFree without a cap, open sourceAbout 10 seconds in the developer exampleYesNo paid versionDevelopers comfortable installing Python software

Try it with your own voice now: record a short sample online, or download Dictator for Windows and macOS. Both free options let you test the result before buying anything.

Which cloning approach fits your task?

Find your task below.

Dictator: unlimited voice cloning on your computer

Dictator (dictator.my) is a Windows and macOS application for text to speech, transcription and voice typing. To create a local clone, record 20-30 seconds of calm speech, download a suitable model, and let the application read your text in that voice. Your own hardware generates the audio, so there is no per-minute cloud bill on any plan, including Free.

Your own voices in Dictator: My voice in the voice list, with buttons to record another voice and create a voice from a description
Your own voices in Dictator: My voice in the voice list, with buttons to record another voice and create a voice from a description

Local cloning does not need an account or a paid plan. Download Dictator free.

ElevenLabs: a familiar option, but no free cloning plan

ElevenLabs appears in many voice-cloning comparisons. In the October 9, 2026 check, its free plan did not include a clone. Instant cloning started with Starter at $6/month, with roughly 30 minutes of synthesis, while Professional cloning started with Creator at $22. For an instant clone, the service recommended 1-2 minutes of clean speech. Requests from Russia redirected to a page about restricted countries in that check.

Best suited to: work for international platforms when you can access the service and use its supported payment methods.

iVox Studio: a Russian subscription service

iVox Studio offered more than 500 text-to-speech voices and cloning from a recording of at least 30 seconds. Registration provided 300 trial characters, while cloning was part of a subscription starting at RUB 490/month for 8,000 characters. Unused characters rolled over to the following month. The website displayed a 50% discount on October 9. Only your own voice could be cloned, with samples subject to moderation.

Best suited to: users without a GPU who need a subscription paid in rubles.

KikiVoice: a trial without registration

KikiVoice works in the browser: upload or record a sample, enter text and generate speech. In the original check, free credits refreshed weekly, and each request accepted 500-2,000 characters depending on the model. No account or payment card was required. We did not find a stated sample length or paid-plan prices on its website.

Best suited to: a short, one-off text in your own voice.

Chatterbox Multilingual: an open model for developers

Chatterbox Multilingual is Resemble AI's open model under the MIT licence. It clones voices in 23 languages, including Russian, and runs on your computer. Installation uses Python and the command line. The developer example uses a sample of roughly 10 seconds. The same model is available among Dictator's cloning models, where you install it with a button rather than configuring the command line.

Best suited to: developers adding voice cloning to their own software.

Why free cloud cloning runs out quickly

Every second of cloud synthesis consumes the service's GPU resources. Its free allowance is therefore usually a trial: 300 characters, weekly credits, or 30 seconds a day. Five minutes of speech is roughly 3,600 characters at a reading pace of 12 characters per second; a complete video does not fit into that kind of trial. The sample also has to be processed by another party's server.

A local program uses your hardware instead. No service is selling you those local minutes, and the sample stays with you. The trade-off is a model weighing several gigabytes that must be downloaded once. Of the six options in the original comparison, Dictator and Chatterbox were the two free local choices without a usage cap. Dictator provides the application interface for people who do not want to set up Python.

Narrate a video in your own voice with Dictator in 3 steps

  1. Download Dictator for Windows or macOS and install it. The local path requires no registration.
  2. In "Text to speech," choose "On this computer," click the action to record your own voice, and record 20-30 seconds of speech. The application shows compatible models and downloads the selected one.
  3. Paste your video script, choose your voice and generate the speech. Download the result as MP3, WAV or OGG.
Dictator's online voice-cloning tab: 30 seconds of generated audio per day, free and without registration
Dictator's online voice-cloning tab: 30 seconds of generated audio per day, free and without registration

How long should the sample be, and what should you read?

Services ask for different sample lengths. The original comparison used 20-30 seconds for Dictator's local flow, at least 30 seconds for iVox Studio, and ElevenLabs' recommendation of 1-2 minutes. Dictator's local recording flow expects at least 20 seconds, because a very short sample is less useful for reproducing the voice. Once the sample is long enough, recording quality matters more.

Can you clone someone else's voice?

Clone your own voice, or a voice you have permission to use. Dictator's online clone asks you to confirm before recording that the voice is yours or that the speaker has permitted its use.

Frequently asked questions

Can I clone a voice free without limits?

The unlimited options in this comparison run on your computer: Dictator, or an open model such as Chatterbox. Online services offer a trial, for example 30 seconds a day, 300 characters, or weekly credits.

Can voice cloning work offline?

Yes. Dictator downloads a model once, then generates speech offline without uploading the sample. You need compatible hardware; the model selector checks the current requirements. The original comparison used an NVIDIA GPU on Windows or an Apple Silicon Mac with at least 8 GB of memory.

Can an AI clone speak Russian?

Yes. Dictator's October catalogue had four cloning models supporting Russian, and its online demo also reads Russian text. ElevenLabs, iVox Studio and Chatterbox Multilingual supported Russian in the comparison.

How many seconds of voice do I need?

Dictator's local flow uses 20-30 seconds; its Premium cloud clone can use a 10-30-second sample. iVox Studio requested at least 30 seconds and ElevenLabs recommended 1-2 minutes in our original check. Record in a quiet room without background music.

Can I clone a voice without registration?

Yes. Dictator's online voice clone works without registration, with 30 seconds of generated audio per day. Its local application flow does not require an account either. KikiVoice provided trial credits without signing in.

Where is my voice stored?

In Dictator's local mode, the sample and clone are stored on your computer. For the online demo, the sample is kept in your browser and sent to the cloud model outside Russia with each synthesis request; the demo does not retain it on Dictator's server. Local and cloud modes therefore have different data paths.

What if I do not have a suitable GPU?

Use a cloud clone. Dictator provides it on Premium at $29.99/month, using the plan's shared 100,000-token balance. In the historical October comparison, iVox Studio offered cloning subscriptions from RUB 490/month.

The simplest test is your own voice: record 30 seconds online, or download Dictator, free.

Competitor limits and prices were checked on their websites on October 9, 2026. Dictator's cloud billing was updated against its code and pricing specification on November 16, 2026. The local-model counts and sizes and Russian legal discussion retain the dates and scope of the original article. Related guides: text to speech with different voices, text to speech, and character voices and online cloning.

Who wrote this

BOZEX develops Dictator (dictator.my), an application for text to speech in your own or library voices, transcription and offline voice typing. Competitor terms were checked October 9, 2026; Dictator's current cloud plans were checked against its code and pricing specification on November 16, 2026.