Features
- Local text-to-speech with BlueTTS
- Fully offline generation after the model is downloaded
- Saved voices: Noa, Lily, Daniel, and Adam
- Supported languages: Hebrew, English, Spanish, German, and Italian
- Audio preview after creation
- 💻 Desktop apps for
macOS on Apple Silicon, Linux on x86_64, and Windows on x86_64
- Local HTTP API with Swagger docs for tools and automation
- Agent-ready
/skill instructions for AI workflows
Install
One command per platform. Each script is plain text at the URL it is fetched
from, so you can read it first if you would rather see what it does before it
does it.
macOS on Apple Silicon
curl -fsSL https://github.com/maxmelichov/MamboTTS/releases/latest/download/install.sh | sh
This downloads the DMG, copies MamboTTS.app into /Applications, and then
removes the quarantine attribute with xattr -dr com.apple.quarantine. That
last step is the reason the script exists. The macOS build is ad-hoc signed
rather than Developer ID signed, because Developer ID signing needs a paid
Apple Developer account, so Gatekeeper refuses the first launch and tells you
nothing useful about why. Clearing the quarantine flag is the same approval as
right-clicking the app and choosing Open, granted once, out in the open, to an
app you asked for by name.
Manual install from the DMG
If you open MamboTTS_<version>_aarch64.dmg yourself and drag MamboTTS.app
into /Applications instead of running the installer above, the app will
refuse to open and Finder will not tell you why. Dragging out of the mounted
DMG keeps the quarantine attribute the installer would otherwise clear, so
clear it yourself:
xattr -dr com.apple.quarantine /Applications/MamboTTS.app
If it still will not open after that, the bundle's ad-hoc signature was never
sealed over its resources (codesign -dvvv /Applications/MamboTTS.app shows
Sealed Resources=none when this is the problem). Re-apply the signature
over the whole bundle:
codesign --force --deep --sign - /Applications/MamboTTS.app
This is exactly what the one-line installer does for you, which is why it is
the recommended path.
Linux on x86_64
curl -fsSL https://github.com/maxmelichov/MamboTTS/releases/latest/download/install.sh | sh
This puts the AppImage in ~/.local/bin and writes a launcher entry into
~/.local/share/applications, so MamboTTS appears in your applications menu.
Nothing here asks for sudo. If your distribution has dropped libfuse2 the
AppImage will complain, and running it with --appimage-extract-and-run gets
around that without installing anything.
Debian, Ubuntu, Fedora, and RHEL users who would rather use their package
manager can take the .deb or .rpm from the
release instead.
Windows on x86_64
irm https://github.com/maxmelichov/MamboTTS/releases/latest/download/install.ps1 | iex
This downloads the desktop installer and runs it. The NSIS window is left
visible on purpose: the command above is pasted from the internet, and the
installer is the one place you get to see where the app is going and say no.
Pass -Silent if you are installing unattended.
The app installs for the current user, so nothing asks for an administrator.
On first launch the onboarding screen downloads the BlueTTS model and the
Renikud phonemizer, around 580 MB, into the app data folder. That is the only
part that needs the network; generation after it is offline.
If you want the HTTP API without the window, -Server installs the headless
server into %LOCALAPPDATA%\MamboTTS with a MamboTTS-Server.cmd launcher
that starts it and opens the Swagger page. From there /docs is a working
console over /v1/audio/speech, /v1/voices, /v1/phonemize, and /skill,
which is everything the desktop app uses to make audio.
Models and phonemizers
MamboTTS builds on these open-source projects:
- BlueTTS is the local ONNX text-to-speech runtime, shipped by default
- RenikudPlus handles Hebrew grapheme-to-IPA conversion with speaker conditioning, and its predictions also add niqqud (diacritics, with the hatama stress mark) to Hebrew text
Python
crates/mambotts-py exposes the same engine to Python and ships a FastAPI
server over it, so the HTTP API is available on the platforms the desktop app
does not ship an installer for, such as Intel macOS and Arm Linux. See its
README.
Build
See BUILDING.md.
Adding models
Want to ship another open-source TTS model or voice bundle with MamboTTS? See docs/ADDING_MODELS.md for licensing, registry wiring, server/desktop steps, and the PR checklist.
This code is taken from Chirp.
