Select and read aloud text from anywhere 🔊
- ffmpeg
- wl-clipboard (Wayland only)
- xclip (X11 only)
- piper C++ (https://github.com/rhasspy/piper/releases), or piper python (https://pypi.org/project/piper-tts/)
- anything to send requests
- Install piper in your $PATH if using the original C++ variant, or
pip install piper-ttsif using the Python wrapper - Download the models and their respective configurations in a directory. See here
- Create a virtual environment, install requirements and run:
python -m venv venv --system-site-packages source venv/bin/activate pip install -r requirements.txt python main.py --port 5000 --speed=1.0 --volume=.8 --piper-model yourmodel.onnx --piper-model-config yourmodel.onnx.json --wayland - Select any text in any application 4. To read aloud:
curl https://localhost:5000/read
- To read aloud random text, send a POST request:
echo Hope you are having a lovely day, sir. | curl -X POST -H 'Content-Type: application/octet-stream' --data-binary @- localhost:5000/read
- To just download the generated audio, instead of playing it:
curl 'https://localhost:5000/read?getaudio' - To interrupt the reading:
curl https://localhost:5000/reset
- To get basic runtime stats:
curl https://localhost:5000/status
- You can dynamically alter the speed and volume using:
curl https://localhost:5000/speed/1.25 curl https://localhost:5000/volume/0.7
- Pause, play, toggle and skip with:
curl https://localhost:5000/pause curl https://localhost:5000/play curl https://localhost:5000/toggle curl https://localhost:5000/skip
- To ignore certain characters in the text:
This will remove all instances of these characters from the text before processing it. You can specify any characters you want to ignore by passing them as arguments after
python main.py --ignore_chars '*' '-'
ignore_chars.
For practical usage, you can set keybindings in your DE or window manager. Say, if you're running sway, add the following to your config ~/.config/sway/config.
bindsym $mod+t exec "curl https://localhost:5000/read"
bindsym $mod+shift+t exec "curl https://localhost:5000/reset"
bindsym Shift+XF86AudioPlay exec "curl https://localhost:5000/toggle"
bindsym Shift+XF86AudioNext exec "curl https://localhost:5000/skip"usage: tts-reader [-h] [--ip IP] [--port PORT] [--wayland | --no-wayland]
[--piper-python | --no-piper-python]
[--speechd | --no-speechd] [--volume VOLUME] [--speed SPEED]
[--piper-rate PIPER_RATE]
[--piper-sentence-silence PIPER_SENTENCE_SILENCE]
[--piper-one-sentence | --no-piper-one-sentence]
[--piper-model PIPER_MODEL]
[--piper-model-config PIPER_MODEL_CONFIG]
[--debug | --no-debug]
[--ignore_chars [IGNORE_CHARS ...]]
options:
-h, --help show this help message and exit
--ip IP IP address
--port PORT Port
--wayland, --no-wayland
Assume running under Wayland
--piper-python, --no-piper-python
Attempt to use the piper python module. Has no effect
if a different backend is selected
--speechd, --no-speechd
Use speechd instead of piper. Incomplete
--volume VOLUME Volume [0-1]
--speed SPEED Playback speed [0-10]
--piper-rate PIPER_RATE
Piper: Playback sample rate. More info at https://gith
ub.com/rhasspy/piper/blob/master/TRAINING.md
--piper-sentence-silence PIPER_SENTENCE_SILENCE
Piper: Seconds of silence after each sentence
--piper-one-sentence, --no-piper-one-sentence
Piper: Process one sentence at a time, instead of the
default whole selection
--piper-model PIPER_MODEL
Piper: Path to the model
--piper-model-config PIPER_MODEL_CONFIG
Piper: Path to the model configuration
--debug, --no-debug Enable flask debug mode (developmental purposes)
--ignore_chars [IGNORE_CHARS ...]
List of characters to ignore