---
title: "Clone My Voice on a Mac: Step-by-Step Guide"
description: "Clone your voice on a Mac with a consent-first, step-by-step workflow: record a clean sample, choose a local model, and check the result before a long script."
canonical: "https://www.murmurtts.com/blog/clone-my-voice-mac-guide"
---
[Murmur](https://www.murmurtts.com/)/[Blog](https://www.murmurtts.com/blog)/Tutorial

Tutorial

# Clone My Voice on a Mac: Step-by-Step Guide

Clone your voice on a Mac with a consent-first, step-by-step workflow: record a clean sample, choose a local model, and check the result before a long script.

![](https://www.murmurtts.com/murmur-icon.webp)Murmur·October 29, 2026·9 min read[Buy $49](https://murmur-licenses.tarunyadav9761.workers.dev/checkout)[Download free for Apple Silicon (3-day Pro trial)](https://murmur-updates.tarunyadav9761.workers.dev/download/latest)Intel Mac? [Download here](https://murmur-updates.tarunyadav9761.workers.dev/download/intel).Not sure? Apple menu > About This Mac shows your chip.

Free plan with Kokoro voices, scripts up to 1,500 characters · 3-day Pro trial, then $49 once

On this page

What cloning needs before you recordStep 1: Confirm consent and set the scopeStep 2: Record a usable referenceStep 3: Choose a local cloning model on the MacStep 4: Import the sample and generate a testStep 5: Check the result before a long scriptWhen Murmur is not the right choiceFrequently asked questions

[Back to the journal](https://www.murmurtts.com/blog)

**Short answer:** You can clone your voice on a Mac by recording a clean sample of one speaker, confirming you have the right to use that voice, importing the recording into a model that supports reference-audio cloning, and testing a single short line before you produce a whole script. The recording and the consent matter more than any setting you change later. This guide walks the Mac steps in order, from permission to a checked result.

It is written for a local workflow on your Mac, where the reference audio stays on your machine instead of being uploaded to a cloud generation service. Every entry-level Mac microphone can produce a usable sample if the room is controlled and the speaking distance is steady.

## What cloning needs before you record

Cloning needs three things that have nothing to do with a settings slider: a clean recording of one speaker, a model that actually supports cloning, and a clear right to use the voice. Murmur is a local voice studio for Apple Silicon and Intel Macs running macOS 15 or later. Its bundled Kokoro model uses preset voices and does not clone a speaker from a recording. The optional cloning models need at least 16 GB of unified memory plus a model download, so check the model list and installation state in the app before you plan a project. If you are new to the topic, the [short-sample cloning guide](https://www.murmurtts.com/blog/clone-your-voice-10-seconds) explains what a reference recording can and cannot promise.

The equipment is the easy part. A built-in Mac microphone, a USB microphone, or even a phone can capture a usable reference if the room is quiet and the microphone distance does not move. The difficult part is judgment: which take to keep, how long the sample should be, and whether the person on the recording agreed to the use you have planned.

## Step 1: Confirm consent and set the scope

Start with permission, not the microphone. If the voice is your own, you already hold the rights, but write down the planned use anyway so later revisions do not drift. If the voice belongs to someone else, get explicit, informed consent before recording a single word. Consent for a voice is separate from a software license: a model being free to use does not give you the right to imitate a person. Keep the reference recording, the transcript, and the consent record together so the scope travels with the audio.

A useful consent note is specific. It should name the project, the platforms where the audio will appear, how long the permission lasts, whether new scripts may be generated from the voice, and how the speaker can ask for deletion. Familiarity is not permission. A colleague, a podcast guest, or a voice heard in a public video has not agreed to be cloned. When the intended use is unclear, write the scope down. The [private voice cloning guide](https://www.murmurtts.com/blog/private-voice-cloning-guide) walks through a fuller consent and storage checklist.

## Step 2: Record a usable reference

Record in a small, softly furnished room with the fan off and notifications silenced. Place the microphone a fixed distance from the mouth, roughly a hand span for a desktop mic, speak at a natural and steady level, and watch the input meter so the loudest line keeps headroom instead of clipping. Clipped peaks cannot be repaired later, so a slightly quiet clean take beats a loud damaged one.

1. Record ten seconds of room tone first and listen on headphones. If you hear a hum, a fan, or traffic, fix the room before recording speech.
2. Read a passage with varied sentence length, not one shouted slogan. Ordinary sentences with normal punctuation represent the voice better than a dramatic line.
3. Keep one speaker on the track. A second voice, music, or a television in the background can be learned and copied by the model.
4. Save an uncompressed WAV master and an exact transcript beside it. Several models reject a reference whose transcript does not match the recording.
5. Trim long silences at the start and end, but leave natural breaths. Skip heavy denoising unless the noise makes the file unusable.

The exact length depends on the model, so capture 30 to 60 seconds and keep a clean 10 to 30 second excerpt. Longer is not automatically better; a short, phonetically varied clip can outperform a long rambling one, and merging recordings from different days can confuse identity when the room or your voice changed. For the full recording workflow, including a file naming worksheet, see the [voice cloning sample recording guide](https://www.murmurtts.com/blog/voice-cloning-sample-recording-guide).

## Step 3: Choose a local cloning model on the Mac

The model you choose decides the reference length, the language coverage, and whether you need a transcript. The table below summarizes the clone-capable models Murmur can install, using the reference guidance shown in the app. Preset-voice models such as Kokoro are not listed because they do not clone a speaker from audio.

| Model | Languages | Reference length guidance | Transcript required |
| --- | --- | --- | --- |
| Chatterbox Turbo | English | More than 5 seconds; 5 to 15 seconds works best | No |
| Chatterbox Multilingual | 23 languages, including English, German, French, Spanish, Japanese, and Chinese | About 3 to 15 seconds | No |
| Qwen3-TTS Base | English, Chinese, Japanese, Korean | 3 to 10 seconds, up to 210 seconds | Yes |
| SparkTTS | English, Chinese | 5 to 15 seconds | Yes |
| Fish Audio S2 Pro | 11 languages, including English, Japanese, Chinese, Spanish, and German | 10 to 30 seconds, up to 210 seconds | Usually no; the 8-bit Mac variant needs it |
| Breeze TTS 2 | English, Chinese | 5 to 15 seconds, up to 60 seconds | Yes |
| OmniVoice | Extended language list, including English, German, French, Spanish, Arabic, and Hindi | 3 to 10 seconds | Yes |

Two patterns are worth noting. First, several models need a transcript that matches the reference exactly, so a recording without a saved script may be rejected before generation begins. Second, an English-only model can clone quickly but will not cover a multilingual script, so match the model to the language you actually need. If the task is designing a brand-new voice rather than copying a real one, that is a different feature: the [cloning, design, and conversion guide](https://www.murmurtts.com/blog/voice-cloning-vs-voice-design-vs-conversion) explains the difference. Keeping these engines in one app means you can switch models without building a separate command-line environment for each one.

## Step 4: Import the sample and generate a test

1. Open the app, choose a cloning model from the list, and let the download finish. The optional models are large compared with the bundled voice.
2. Add the reference through the reference-voice workflow, then attach the exact transcript if the model requires one.
3. Write a short test sentence that includes a normal statement, a question, a name, and a number. A name and a number expose pronunciation errors that a plain sentence hides.
4. Generate that one line first. If it fails, change one variable at a time: try a cleaner reference, fix a transcript mismatch, or shorten an ambiguous clause.
5. Only after the test sounds right, paste the real script and generate it in sections rather than as one enormous file.

Keep the untouched master recording even after the clone works. A voice that fits today may need a better reference for the next project, and re-recording a speaker is harder than keeping the original file.

## Step 5: Check the result before a long script

Judge the output in two passes. First, check that the words are complete: no dropped endings, no repeated phrases, and correct numbers and names. Second, compare speaker similarity against the original at a similar volume, without telling a listener which take is which. Repeat a few times, because a convincing result that happens once is not yet a reliable production voice.

- Script completion: every sentence, name, and number is present.
- Speaker similarity: the identity holds across several sentences, not just the first one.
- Pronunciation: proper nouns and technical terms sound the way you expect.
- Stability: three runs do not drift in pace, pitch, or accent.
- Delivery: the tone suits the piece, not just the demo sentence.

If repeated repairs make the result worse instead of better, stop. A different model, a preset voice, or a human recording may be the faster answer, and that is a normal outcome rather than a failure. Do not promise a client an identical voice based on one short test.

## When Murmur is not the right choice

Murmur is a Mac app, so it is not an option on Windows, and there is no iPhone, Android, or browser version. It does not do live or real-time streaming speech, and it does not offer cloud collaboration where several people edit one project at once. It also does not do speech to text, so it cannot transcribe an interview into a script. If you need any of those, a cloud service that runs in a browser on any device is the better fit.

Cloud tools are also strong on convenience: a shared voice library, team accounts, and no model download. The tradeoff is that your script and reference audio may leave your machine, and per-character fees grow on long projects. Local generation suits private or client work, but local also means you supply the disk space and the memory. For a side-by-side, read the [local voice cloning apps for Mac](https://www.murmurtts.com/blog/best-local-ai-voice-cloning-apps-mac) comparison. If you only want to try one cloned sentence with no setup at all, decide on the public samples first: Murmur has a free plan and a 3-day Pro trial, costs $49 one time, and offers a 7-day refund policy.

## Frequently asked questions

## Sources

- [Resemble AI: Chatterbox official repository and reference-audio usage](https://github.com/resemble-ai/chatterbox)Accessed 2026-10-14
- [Qwen: Qwen3-TTS-12Hz-0.6B-Base official model card (3-second voice cloning)](https://huggingface.co/Qwen/Qwen3-TTS-12Hz-0.6B-Base)Accessed 2026-10-14
- [Apple: Create a recording on Mac with Voice Memos](https://support.apple.com/guide/voice-memos/make-a-recording-vmaa4b813415/mac)Accessed 2026-10-14

## Take a consented sample to a local clone

Murmur is a local voice studio for Apple Silicon and Intel Macs running macOS 15 or later. It keeps reference voices, scripts, and exports on your machine for $49 one time, with a free plan and a 3-day Pro trial, and a 7-day refund policy. Listen to the public samples before you buy. Murmur is $49 one-time for Apple Silicon and Intel Macs on macOS 15 or later, with a free plan and a 3-day Pro trial, and a 7-day refund policy.

[Buy Murmur · $49](https://murmur-licenses.tarunyadav9761.workers.dev/checkout)[Download free for Apple Silicon (3-day Pro trial)](https://murmur-updates.tarunyadav9761.workers.dev/download/latest)Intel Mac? [Download here](https://murmur-updates.tarunyadav9761.workers.dev/download/intel).Not sure? Apple menu > About This Mac shows your chip.

macOS 15+ · Apple Silicon and Intel Macs · Windows beta · 7-day refund policy

Made for your Mac

## Less reading. More making.

[Buy $49](https://murmur-licenses.tarunyadav9761.workers.dev/checkout)[Download free for Apple Silicon (3-day Pro trial)](https://murmur-updates.tarunyadav9761.workers.dev/download/latest)Intel Mac? [Download here](https://murmur-updates.tarunyadav9761.workers.dev/download/intel).Not sure? Apple menu > About This Mac shows your chip.[Explore Murmur](https://www.murmurtts.com/#product)