Text to speech · Voice cloning · Voice design · Multi-speaker projects · CLI + MCP
Download · Hear samples · Website · Documentation
Murmur is a native macOS app for turning scripts into speech, creating reusable voices, and producing multi-speaker audio. Speech generation runs locally on your Mac. Download models once, then generate without a cloud speech API or per-character fees.
This is Murmur’s public documentation, changelog, and issue tracker. The commercial Mac application’s source code is not included here. See LICENSE for the scope of this repository’s documentation license.
- Create speech: write or import a script, choose a compatible voice and model, adjust supported controls, and export WAV or M4A.
- Clone a voice: import or record a reference sample, review its transcript, and save a reusable profile. Sample requirements vary by model; Murmur shows guidance before generation.
- Design a voice: describe a narrator or character in plain language, browse 50 prompt presets, preview the result, and save it to My voices.
- Produce a project: organize scripts and speakers, assign voices, generate dialogue, arrange audio on a timeline, and add music or sound effects before export.
- Manage voices: search designed and cloned profiles, or browse and preview community voices in Explore.
- Queue work: add multiple scripts and generate them as a batch. Import DRM-free EPUB books for chapter-based narration.
- Automate locally: generate speech from Terminal or connect Codex, Claude Code, and Cursor through Murmur’s MCP server.
These are screenshots of the native Mac app, not website mockups. Voice names, history, model availability, and playback state reflect the example workspace shown.
Scripts, assigned speakers, dialogue editing, a media timeline, and an inspector in one workspace.
Describe the voice, browse the prompt library, and generate a preview before saving.
Designed and cloned voices appear in separate groups, with search, filters, and reusable voice selection.
Browse community profiles, filter by traits and language, listen to samples, and choose a voice. The catalog changes over time; the screenshot shows 984 voices.
Set generation and export defaults, manage models and local storage, enable automation, and check for updates.
Kokoro is bundled. Other models are optional local downloads managed in Settings → Models. Controls and language support depend on the selected model.
| Model | Main capabilities |
|---|---|
| Kokoro | Lightweight speech generation with built-in voices, language selection, and speed control |
| Chatterbox Turbo | English voice cloning and expressive speech tags |
| Chatterbox Multilingual | Voice cloning across 23 languages |
| Qwen3-TTS Base | Preset voices, reference voice cloning, and multilingual speech |
| Voice Design (Qwen3-TTS) | Generate a custom voice from a natural-language description |
| SparkTTS | Reference voice cloning for English and Chinese |
| Fish Audio S2 Pro | Expressive voice cloning, expression tags, and speed control; reference samples require an exact transcript |
| OmniVoice | Omnilingual reference cloning, with support for 600+ languages in the model |
Compare real samples · Language guide · Browse voices
- Download Murmur and move it to Applications.
- Open the app and complete setup and license activation.
- Use bundled Kokoro, or install an optional model in Settings → Models.
- Open Create speech, enter your text, choose a voice, and select Generate.
- Listen to the result and export it as WAV or M4A.
For longer productions, open Projects to organize scripts, assign speakers, and work with a timeline.
- macOS 15 or later
- Apple Silicon (M-series Mac); Intel Macs are not supported
- Free storage for the app, generated audio, and any optional models you choose
- Enough unified memory for the selected model; larger models require more resources
- Internet access for initial setup, model downloads, license activation, updates, and community browsing; speech generation runs locally after setup
Model installation manages its own local runtime. Homebrew, Xcode Command Line Tools, and administrator access are not required for optional speech runtimes.
Enable Settings → Automation → Allow local automation, then install the CLI from that screen. Keep Murmur installed in Applications.
murmur status --json
murmur models --json
murmur voices --model kokoro --json
murmur generate --text "Hello from Murmur." --model kokoro --output hello.wavTo use your own script:
murmur generate --input narration.txt --model kokoro --output narration.wavConnect an AI coding agent:
# Codex
codex mcp add murmur -- /Applications/Murmur.app/Contents/Helpers/murmur mcp serve
# Claude Code
claude mcp add --transport stdio --scope user murmur -- /Applications/Murmur.app/Contents/Helpers/murmur mcp serveFor Cursor, add this to your project’s .cursor/mcp.json:
{
"mcpServers": {
"murmur": {
"command": "/Applications/Murmur.app/Contents/Helpers/murmur",
"args": ["mcp", "serve"]
}
}
}MCP supports model and voice discovery, speech generation, batches, job progress, and cancellation. Integrations can read and write files only within their starting workspace. Existing outputs are not replaced unless replacement is explicitly requested.
Speech synthesis and reference-voice processing happen on your Mac. Local generation does not upload your scripts or reference audio to a hosted speech service. Model downloads, community browsing, licensing, and updates use network connections.
Use voices and source material you have permission to use. Commercial use remains subject to Murmur’s terms, applicable model licenses, and rights to the reference voice and content.
Privacy policy · Terms · Security policy
Murmur is a one-time purchase, with no subscription or per-character generation charge. See the website for current pricing and the refund policy.
- Report a bug
- Request a feature
- Email tarunyadav9761@gmail.com
- Release notes
- Projects guide · Voice Design · Use cases
Built by Tarun Yadav.