Re: Any Progress on an AI Related Screen-Reader?
I'm pretty sure that Samuel was talking about piper-tts.
After running `pip install piper-tts` you first have to install a voice
model. You can hear samples of the voices here:
https://rhasspy.github.io/piper-samples/
The easiest way to actually download a voice model would be by using the
piper.download_voices program, which you run like this:
$ python3 -m piper.download_voices
This will list the available voices. To filter for a particular
language, you can use grep:
$ python3 -m piper.download_voices | grep '^en_'
Then use the same program to download the model you want by passing it
the model name:
$ python3 -m piper.download_voices en_US-arctic-medium
Finally, you can test it using the piper command:
$ echo "Hello world!" | piper -m en_US-arctic-medium --output-raw |
aplay -r 22050 -f S16_LE
A model can contain multiple speakers, so you might want to use
something like:
$ echo "Hello world!" | piper -m en_US-arctic-medium -s 2 --output-raw |
aplay -r 22050 -f S16_LE
I hope that helps.
On 8/7/25 14:18, Chime Hart wrote:
Hi Jean: I am not a programmer nor developer, just a happy Debian SID blind
user who would like additional screen-reader-and-voice options. My current
Speakup screen-reader doesn't have an option to announce numbers as
single-digits. Since I have well over 11thousand messages in my inbox, it would
certainly save time hearing an index as single digits. Most early DOS
screen-readers had this-and-other options. After Samuel's suggestion, I ran
pipx install piper
but its not obvious how to run it. Not even any man-page with a word "piper"
Thanks in advance
Chime
Reply to: