zenity script for stdout of piper text to speech

[Coffeeee]

I was looking for a good tts and found piper tts.
You can hear voice samples here
https://rhasspy.github.io/piper-samples/

Easiest to use the binary
https://github.com/rhasspy/piper#installation

Needs jq installed
sudo apt install jq

put this script in the same directory as the binary

Code: Select all

#!/bin/bash

cd "$(dirname "$0")"

while true; do

TEXT_IN=$( zenity --entry --title="piper_tts" --text="enter/paste text to read" --cancel-label=exit --ok-label=read --extra-button "read clipboard" --extra-button pause --extra-button resume --extra-button kill)

if [ $? = 0 ] # check if the user click ok on the zenity form
then
	echo "$TEXT_IN" | ./piper --model en_GB-vctk-medium.onnx --length_scale 1.9 --noise_scale 0.333 --noise_w 0.33 --output-raw | aplay -r 20000 -f S16_LE -t raw - &
  elif [[ $TEXT_IN = "read clipboard" ]]
   then
		echo $(xclip -o) | ./piper --model en_GB-vctk-medium.onnx --length_scale 1.9 --noise_scale 0.333 --noise_w 0.33 --output-raw | aplay -r 20000 -f S16_LE -t raw - &
  elif [[ $TEXT_IN = "pause" ]]
   then
        pid=$(pidof piper); kill -STOP $pid;
  elif [[ $TEXT_IN = "resume" ]]
   then
        pid=$(pidof piper); kill -CONT $pid;
  elif [[ $TEXT_IN = "kill" ]]
   then
        killall piper;
else
    exit
fi

done

#    x_low - 16Khz audio, 5-7M params
#    low - 16Khz audio, 15-20M params
#    medium - 22.05Khz audio, 15-20 params
#    high - 22.05Khz audio, 28-32M params

if you understand how to use wmctrl then use below (keep window on top, set position)

Code: Select all

#!/bin/bash

cd "$(dirname "$0")"

while true; do

sleep 0.4 &&
wmctrl -F -a "piper_tts" -b add,above &&
wmctrl -r "piper_tts" -e 0,2300,90,100,100 &
TEXT_IN=$( zenity --entry --title="piper_tts" --text="enter/paste text to read" --cancel-label=exit --ok-label=read --extra-button "read clipboard" --extra-button pause --extra-button resume --extra-button kill)

if [ $? = 0 ] # check if the user click ok on the zenity form
then
	echo "$TEXT_IN" | ./piper --model en_GB-vctk-medium.onnx --length_scale 1.9 --noise_scale 0.333 --noise_w 0.33 --output-raw | aplay -r 20000 -f S16_LE -t raw - &
  elif [[ $TEXT_IN = "read clipboard" ]]
   then
#		xclip -o > ~/piper.txt
#		sed -i 's/~/-/g' ~/piper.txt
#		cat ~/piper.txt | xclip;
		echo $(xclip -o) | ./piper --model en_GB-vctk-medium.onnx --length_scale 1.9 --noise_scale 0.333 --noise_w 0.33 --output-raw | aplay -r 20000 -f S16_LE -t raw - &
  elif [[ $TEXT_IN = "pause" ]]
   then
        pid=$(pidof piper); kill -STOP $pid;
  elif [[ $TEXT_IN = "resume" ]]
   then
        pid=$(pidof piper); kill -CONT $pid;
  elif [[ $TEXT_IN = "kill" ]]
   then
        killall piper;
else
    exit
fi

done

#    x_low - 16Khz audio, 5-7M params
#    low - 16Khz audio, 15-20M params
#    medium - 22.05Khz audio, 15-20 params
#    high - 22.05Khz audio, 28-32M params

I did try the Speech Note flatpak but it writes all output to wav files wearing the disk. Flatpaks also don't work with dark themes.

[Coffeeee]

I just found out how to alter the speed

echo "$TEXT_IN" | ./piper --model en_GB-vctk-medium.onnx --length_scale 1.9 --noise_scale 0.333 --noise_w 0.33 --output-raw | aplay -r 22050 -f S16_LE -t raw -

you can get this information inside the json file matching the voice (onnx)

# --noise_scale NUM generator noise (default: 0.667)
# --length_scale NUM phoneme length (default: 1.0), this controls speed.
# --noise_w NUM phoneme width noise (default: 0.
# --sentence_silence NUM seconds of silence after each sentence (default: 0.2)

[Coffeeee]

accidental post, updated anyway.

Code: Select all

#!/bin/bash

cd "$(dirname "$0")"

## piper json settings
model=en_US-libritts_r-medium.onnx						### MUST BE SET BY USER ###
config=$model.json
length_scale=$(grep length_scale $config | cut -d , -f1 | awk '{print $2}')
frequency=$(grep sample_rate $config | cut -d , -f1 | awk '{print $2}')

## override piper json settings
#length_scale=1.35	# adjusts reading speed. start between 1-2		### ADJUST FOR COMFORT ###
#frequency=20000	# adjusts audio frequency					### ADJUST FOR COMFORT ###
					# x_low & low quality model uses 16Khz audio
					# medium & high quality 22.05Khz
					# adjust for comfort e.g. 22050 at 20000 lowers pitch and speed.

## window position
vertical=90									### MUST BE SET BY USER ###
horizontal=2300								### MUST BE SET BY USER ###

while true; do

sleep 0.5 &&
wmctrl -F -a "piper_tts" -b add,above &&
wmctrl -r "piper_tts" -e 0,$horizontal,$vertical,100,100 &
TEXT_IN=$( zenity --entry --title="piper_tts" --text="enter/paste text to read" --cancel-label="read clipboard" --ok-label=read --extra-button pause --extra-button kill --extra-button clip2wav --extra-button exit)

if [ $? = 0 ] # check if the user click ok on the zenity form
then
	echo "$TEXT_IN" | ./piper --model $model --config $config --length_scale $length_scale --output-raw | aplay -r $frequency -f S16_LE -t raw - &

	elif [[ $TEXT_IN = "pause" ]]
	 then
	  pid=$(pidof piper); kill -STOP $pid;
	  sleep 0.5 &&
	  wmctrl -F -a "piper_tts_resume" -b add,above &&
	  wmctrl -r "piper_tts_resume" -e 0,$horizontal,$vertical,100,100 &
	  zenity --question --title="piper_tts_resume" --text="Piper is paused.\nResume piper?" --cancel-label=kill --ok-label=resume

		if [ $? = 0 ] # check if the user click ok on the zenity form
		 then
		  pid=$(pidof piper); kill -CONT $pid
		 else
		  killall piper;
		fi

  	elif [[ $TEXT_IN = "kill" ]]
  	 then
	  killall piper;
	  
	elif [[ $TEXT_IN = "clip2wav" ]]
  	 then
  	 	fix_clip=$(xclip -o -sel clip | sed -e 's/-chan/-chun/g' -e 's/ LV/ level /ig' -e 's/LV/ level /g' -e 's/-senpai/-senpie/g' -e 's/pyon/p-yon/ig' -e 's/-kun/-/g' -e 's/~/-/g')
		fname=$(zenity --file-selection --save --title="Save wav" --confirm-overwrite --filename="*.wav")
			echo "$fix_clip" | ./piper --model $model --config $config  --length_scale $length_scale  --output_file "$fname"
	elif [[ $TEXT_IN = "exit" ]]
	 then
	  exit
		else
		xclip -o -sel clip | sed -e 's/-chan/-chun/g' \
		-e 's/ LV/ level /ig' \
		-e 's/LV/ level /g' \
		-e 's/-senpai/-senpie/g' \
		-e 's/pyon/p-yon/ig' \
		-e 's/-kun/-/g' \
		-e 's/~/-/g' | ./piper --model $model --config $config  --length_scale $length_scale --output-raw | aplay -r $frequency -f S16_LE -t raw - &
fi

done