Tiktok speech to text parser

tiktok.sh

#!/bin/zsh

# Tiktok speech to text parser
# by hugeblank, April 2022
# API endpoint & oneshot reading and playing discovered by @scanlime
# https://twitter.com/scanlime/status/1512288857596653568

# First argument - string of text to read
# Second argument - voice to use
# Usage examples:
# ./tiktok.sh play "this is a test of my patience" en_us_002
# ./tiktok.sh save "this will get saved to out.mp3" en_us_009
# NOTE: The endpoint appears to have a hard limit of about 300 characters.

# Valid Voices (tested from 0-30 for each language):
# en_us_001 = Standard female
# en_us_002 = Standard female (same as 001 it seems.)
# en_us_006 = British male 1
# en_us_007 = Standard male 1 
# en_us_009 = Standard male 2
# en_us_010 = British male 2

# en_uk_001 = Too British male 1
# en_uk_003 = Too British male 2

# en_au_001 = Australian female 1
# en_au_002 = Australian male 1

# For more, see: https://github.com/oscie57/tiktok-voice#voice-options


str=`echo $2 | tr ' ' '+'` # sanitize spaces
result=`curl -s -X POST 'https''://api16-normal-useast5.us.tiktokv.com/media/api/text/speech/invoke/?text_speaker='$3'&req_text='$str` # make request

statcode=`echo $result | jq .status_code` # make sure request was successful
if [ $statcode -ne 0 ]
then
	echo "could not parse output"
	echo $result
	exit $statcode
fi

output=`echo -En $result | jq .data.v_str | base64 -di` # convert output

if [[ $1 == "play" ]] 
then # play one-off sound
	echo -En $output | ffplay -volume 10 -f mp3 -autoexit -i -
elif [[ $1 == "save" ]] 
then # save to an output (out.mp3)
	rm out.mp3 2>/dev/null
	echo -En $output | ffmpeg -f mp3 -i - out.mp3
else # option one wasn't one of play or save
	echo "Expected first argument to be 'save' or 'play'"
	exit 1
fi

if [ $? -ne 0 ] # either ffplay or ffmpeg errored
then
	# both make it very apparent that they errored, this is effectively moot.
	echo "couldn't decode sound"
	exit $?
fi

thanks a ton for this! i had to install jq for this (understandably), and had to modify line 43 to read as follows – on macOS where base64 works a bit differently, i think:

output=`echo -En $result | jq .data.v_str | tr -d '"' | base64 -d`

A cleaner change to make things work on macOS, where base64 doesn't support the -i option to ignore invalid input characters, is to ask jq not to include them with the -r (--raw-output) flag:

output=`echo -En $result | jq -r .data.v_str | base64 -d` # convert output`

hah, that's a much better solution. thanks @JiBB!

I don't know why I didn't mention this before, but this script broke a while ago. There's a fix that was just pushed to oscie57's tiktok-voice that I'm awaiting permission to implement.