Skip to content
Yummi
PTEN

[ INSTALL ]

Opening Yummi

Four commands and one output choice. The virtual audio cable comes later, when you want other programs to hear you.

Before you start

01

Python 3.10 or newer

It is the only language involved — Yummi is a Python package end to end.

02

Windows, Linux or macOS

Speech, files and the library work on all three. Per-window capture is Windows only, and needs build 20348 or newer.

03

A virtual audio cable (optional)

VB-CABLE is free, and it is what turns Yummi's output into a microphone for other programs.

VB-CABLE ↗

The four commands

On Windows, yummi-gui opens without the black console behind it. yummi keeps the console, which helps when something goes wrong.

  1. 01Download the code.

    git clone https://github.com/ayunaangel/Yummi_text-to-speech.git

  2. 02Enter the folder.

    cd Yummi_text-to-speech

  3. 03Install the package and its five dependencies.

    pip install -e .

  4. 04Open the window.

    yummi

Becoming your microphone

Windows won't let a program write straight into a microphone. The way through is a virtual audio cable: Yummi plays on one end, other programs listen on the other.

VB-CABLE ↗
  1. 01Install VB-CABLE and restart the computer.
  2. 02In Yummi, under OUTPUT (BECOMES YOUR MIC), pick CABLE Input. It is selected by default when detected.
  3. 03Under MONITOR (YOU HEAR), pick your headphones.
  4. 04In Discord, OBS or your game, select CABLE Output as the microphone.

Changing the voice and the rest

The window has no preferences menu on purpose. Everything adjustable lives in a JSON file, in %APPDATA%\Yummi (on Linux, ~/.local/share/Yummi):

To see every available voice, run python -m edge_tts --list-voices and put the chosen name in voz_edge. The default is Francisca, in Brazilian Portuguese.

config.json

{
  "saida_voz": "CABLE Input (VB-Audio Virtual Cable)",
  "saida_monitor": "Fones de ouvido",
  "voz_edge": "pt-BR-FranciscaNeural",
  "idioma_gtts": "pt-br",
  "pasta_audio": "C:\\Users\\voce\\Music\\sons",
  "atalhos_globais": false
}

If something doesn't go as expected

CABLE Input isn't in the output list.
Restart after installing VB-CABLE and press F5 in the Yummi window, which rescans devices.
The BRIDGE tab lists no windows.
Per-process capture exists from Windows 10 build 20348 on. On older versions only audio inputs show up, like stereo mix.
Speech is slow or doesn't come out.
The first time for each line depends on the internet. If edge-tts doesn't answer, gTTS steps in; from the second time on the line comes from cache, with no network.