Skip to content
Built for Apple Silicon

Text-to-speech that sounds human, and stays on your Mac.

Write a script, pick or clone a voice, export WAV. No credits, no subscription, nothing uploaded.

Free to download, Pro free for 3 days · 7-day refund · macOS 15+ on Apple Silicon

MurmurKokoro
Voice Heart Lang English (US) Speed 1.0x Style Fine-tune
Welcome back to the show. Today: voices that never leave your Mac, with no credits and no uploads.
98 chars · 18 words · ~7s Import Queue Generate

Preview of the Mac app. Voices play recordings made in Murmur; generation, cloning and installs are simulated.

Introducing Murmur

Most cloud voice tools bill by the character and process your script on their servers. Every project adds to the bill, and the bill never stops.

Murmur is a voice studio built to run on your Mac. It generates, clones and designs voices locally, so you can make podcasts, audiobooks and voiceovers with no credits, no subscription and no uploads.

$49
One time, for one Mac
0
Credits, tokens or per-word fees
860+
Community voices to start from
600+
Languages with the optional OmniVoice model

Hear it

Press play . Every voice here was made in Murmur.

Some ideas start as a scribble. A sentence in your notes. A story you cannot stop thinking about. Give yours a voice.

0:00 / 0:07

Kokoro · English (US). Pre-recorded in Murmur on a Mac. The app generates new audio from your own scripts.

What you can make

Anything you write, Murmur can say for you.

Podcasters, audiobook authors, YouTubers, course creators, game developers and agencies all start the same way: a script. Murmur runs the voice model on your Mac, so you can make as much as you need. The only limit is the script.

Read my new intro in my own voice

Your recording

my-voice.wav0:00 / 0:12

Murmur, reading something new

Same voice0:00 / 0:09

Clone. Give Murmur a short recording and it reads new scripts in that voice. The reference here is a synthetic Voice Design voice. Needs 16 GB+ memory and an optional model. Starter kit

A warm British storyteller. Hushed, then delighted.

The storytellerEnglish (UK) · Female · ThirtiesHushedWonderIntimate
Preview0:00 / 0:08

Design. Describe who is speaking and how, and Murmur builds a voice from the sentence. No recording needed. Voice Design

Turn this script into a two-host podcast

2 speakers · 1 music bedGenerate all

Produce. Cast every speaker, fix a single line on the timeline, and export WAV or M4A. Projects

Whisper it. Then let the tension break.

whisper Do you hear that? long pause There it is again. inhale Something is moving behind the wall…

Whisper and suspense0:00 / 0:20

Direct the delivery. Write [whisper], [laughing] or [long pause] in the script and expressive models perform it. Optional Fish S2 Pro model. More expressive samples

Narrate this in French. 600+ languages to pick from.

French narration0:00 / 0:06

Many languages. Kokoro covers the common ones. The optional OmniVoice model reaches 600+; coverage varies by model. Compare coverage

Automation

The voice studio your scripts and agents can call.

Run Murmur from Terminal in a build step, or connect an AI agent through MCP. The same local models, the same private audio, no API key to rotate.

  • Batch a whole folder of scripts
  • Works in CI and in shell scripts
  • MCP server for agents
CLI and MCP
  • 860+ community voices

    Start from a voice someone shared. Check each voice's permissions before commercial use.

    Browse the library
  • Batch and queue

    Line up a whole book or a folder of scripts and let your Mac work through it.

    See Projects
  • WAV and M4A export

    Take finished audio straight into your editor, with no watermark and no export limit.

    Export options
  • Works offline

    Once a model is set up, generation needs no connection. License checks and updates still do.

    Privacy details

Explore voices

860+ voices to start from.

Murmur opens a shared library of voices people have made. Listen to any of them, then press Use to narrate your own script. A few, picked from the library:

  • English

    Adrian

    Male · Middle Aged

    A steady and reliable narrator.

  • English

    Sarah

    Female · Young

    An engaged speaker.

  • English

    Ethan

    Male · Middle Aged

    A curious explainer.

  • English

    Sleepless historian

    Eric Adjei

    This is a calm and professional male voice with a clear British accent and a measured tone. It has an informative quality well-suited for educational narration or storytelling.

  • English

    horror

    Modestas

    A deep and serious middle-aged male voice with a calm, mysterious tone. Well-suited for suspenseful narration, documentaries, or cinematic storytelling.

    113.3 K 2.1 KUse
  • Chinese

    温柔动听女声

    412887114

    说话抑扬顿挫,语气温柔动听的女孩子

    46.6 K 1.3 KUse
  • French

    Clémence

    Belkacem Brik

    25+ French Female deep voice, narrator, story, vlog en français

  • Japanese

    元気な女性

    比留間大地

    A youthful and friendly female voice with a gentle, polite tone. Her speech is clear and expressive, making it well-suited for conversational and narrative content.

    115.5 K 554Use

The meter

What are you renting right now?

Most voice tools bill by the character or by the month, and the bill grows with every project. Murmur runs on your Mac, so there is nothing to meter. Move the slider to what you pay today.

See side-by-side comparisons

Murmur pays for itself in 2 months. Over three years that is $1,031 back in your pocket.

Your number, not ours. We do not assume anyone else’s pricing. Taxes are added at checkout, and speed and quality depend on the model and your Mac.

The studio

A script, a cast and a timeline. Not a text box.

  1. 01

    Write or import the script

    Type, paste or drop in a file. Press / to add a speaker, and drop in [pause] or [laugh] wherever you want one.

  2. 02

    Cast every speaker

    Give each speaker a voice of their own: a built-in one, a community voice, a clone of yours, or one you described.

  3. 03

    Fix it on the timeline, then export

    Regenerate a single block, nudge clips around, and export WAV or M4A when it sounds right.

The Murmur Projects window. A two-speaker script with expression tags sits above a timeline of generated Host and Guest clips.
The real app. Not a mockup.

On-device

Your script never leaves the Mac.

Generation runs on your Apple Silicon chip, so there is no upload step and no per-character bill. Here is exactly what stays put and what still needs a connection.

Read the privacy policy

Stays on your Mac

  • Your scripts and every generated clip
  • Cloned voice samples and designed voices
  • Speech generation, once a model is set up
  • Projects, history and exports (WAV or M4A)

Needs the internet

  • Purchase and activation
  • Periodic license validation
  • Model downloads and app updates
  • Browsing community voices
System
macOS 15 or later
Chip
Apple Silicon, M1 or newer. Intel Macs are not supported.
Memory
Bundled Kokoro runs on any compatible Mac. Cloning and Voice Design need 16 GB+ unified memory.
Downloads
Optional models are extra downloads. Setup details

From Murmur customers

In their own words.

“This is perfect, No buying credits, or monthly subscriptions. It just works”
John
“Murmur is an excellent AI voice cloner, I use it a lot to make DJ jingles/drops”
Lucas Schmidt · excerpt
Read the full review

Murmur is an excellent AI voice cloner, I use it a lot to make DJ jingles/drops, and is even great for cloning your own voice if you can’t be bothered to record yourself. I was recently disturbed that on a previous update there seemed to be a 30 second sample limit, thereby breaking all my voice models. For me that rendered it useless, as I really didn’t want to have to start from scratch. I am very pleased to say that has now been fixed and Murmur is best in its class. All my models are now working once more. I thoroughly recommend it!

“Excellent customer service. Easy to use interface. The main model Fish-speech produces high quality audio results. Highly recommended.”
K. Kwatamani · April 11 update excerpt
Read the full review

We received a prompt response regarding our issue and look forward to the fully functioning fish-audio model. (Update April 11: Excellent customer service. Easy to use interface. The main model Fish-speech produces high quality audio results. Highly recommended.)

“Amazing app, incredible quality! And the best part is it is all offline!”
Murmur customer

Pricing

Pay once . Make more .

From a single voiceover to your next audiobook, the production stays on your Mac and the price stays where it is.

  • 7-day money-back guarantee
  • Apple notarized, no account required
Murmur for Mac$49 USD per Mac · one-time
  • One purchase. One license for one Mac.
  • Unlimited local text-to-speech
  • Voice cloning and Voice Design
  • Projects, multi-speaker, batch queue
  • Export WAV or M4A
  • CLI and MCP automation

One license, one Mac.

7-day money-back guarantee. Not sure yet? Download it free: Kokoro stays free, and Pro is free for 3 days. License key by email, taxes at checkout. macOS 15+ on Apple Silicon; cloning and Voice Design need 16 GB+ memory. Intel Macs are not supported.

Before you buy

A few good questions.

Still wondering? Talk to the maker.

What does one purchase include?

Murmur is $49 USD as a one-time purchase. One purchase. One license for one Mac. Generate locally without credits, per-word charges, or a monthly subscription. Applicable taxes are shown at checkout.

Is there a free version?

Yes. Download Murmur and use it free: Kokoro voices, scripts up to 1,500 characters, export to WAV. Pro adds cloning, Voice Design, the other voice models, projects, the batch queue and automation. You can try all of Pro free for 3 days with one click inside the app, with no card, no email and no key to paste. When it ends, the app drops back to the free plan and your projects and audio stay on your Mac. Each Mac gets one Pro trial.

Will it work on my Mac?

You need macOS 15 or later and Apple Silicon (M1 or newer). Intel Macs are not supported. Kokoro handles basic narration on compatible Macs; optional models, including voice cloning and Voice Design, require at least 16 GB of unified memory. Speed and available features depend on the selected model and your hardware.

What do I need to download?

Kokoro is bundled for basic text-to-speech. Optional models for cloning, Voice Design, and other speech features require at least 16 GB of unified memory and additional model/runtime downloads. For example, the app estimates about 1.7 GB for Qwen3-TTS Base and 2.3–4.5 GB for Voice Design, depending on the variant. These are model downloads, not total installation sizes: allow additional free space for runtimes, setup, and generated audio. Total disk space needs and generation speed vary by model and your Mac.

Does it work without the internet?

Speech generation runs locally after setup, but periodic online license validation is still required. Internet is also needed for purchase, activation, optional model downloads, updates, and community voice browsing. Diagnostic connections are described in the Privacy Policy; local generation does not upload your scripts or generated audio. Read the Privacy Policy.

Can I use the audio in commercial projects?

Your Murmur purchase licenses the app. You must also follow the selected speech model’s license and have the rights and permissions needed for the script, reference recording, and voice. Community voices do not all come with the same permissions. Check those terms before publishing or doing client work. Read the usage terms.

What about updates or moving to a new Mac?

Your purchase includes perpetual access to the version you buy and free updates for that version. Future versions and upgrades are not guaranteed. The individual license covers one active Mac; a Mac bundle covers the number purchased. To move to another Mac, the previous activation must be released. Contact support if you need help changing Macs.

What happens after I buy?

Your download and license key arrive by email. Install Murmur, activate your license, and start with the bundled Kokoro model. Add optional models for other workflows. If Murmur does not fit your workflow, email your purchase receipt within 7 days to request a refund.

What does Apple notarization mean?

Apple notarization checks submitted software for malicious components. It is not an App Store review or an endorsement.

Your next voiceover is one script away.

Free to download · 7-day refund · Apple Silicon Macs · Hear a sample again