Keep silence in Jingles

[LordHelmchen666]

I have some Jingles with one second silence at the beginning and the end. How to keep it?
In Azuracast i set fade-in/fade-out/overlap to 0.1, but autocue2 sets cue_in and cue_out.

2024/04/15 16:16:00 [autocue2:3] Autocue2 result for "/var/azuracast/stations/eisradio/media/Promos/randomjingle.mp3": {"duration": "4.60", "liq_duration": "2.90", "liq_cue_in": "1.00", "liq_cue_out": "3.90", "liq_longtail": "false", "liq_cross_duration": "0.40", "liq_loudness": "-15.62 dB", "liq_amplify": "-2.38 dB", "liq_blank_skipped": "false"}

2024/04/15 16:16:00 [autocue2.metadata:3] Metadata added/corrected:
2024/04/15 16:16:00 [autocue2.metadata:3] ("duration", "5.43")
2024/04/15 16:16:00 [autocue2.metadata:3] ("liq_amplify", "-2.39 dB")
2024/04/15 16:16:00 [autocue2.metadata:3] ("liq_blank_skipped", "false")
2024/04/15 16:16:00 [autocue2.metadata:3] ("liq_cross_duration", "0.10")
2024/04/15 16:16:00 [autocue2.metadata:3] ("liq_cue_in", "1.00")
2024/04/15 16:16:00 [autocue2.metadata:3] ("liq_cue_out", "3.90")
2024/04/15 16:16:00 [autocue2.metadata:3] ("liq_duration", "2.90")
2024/04/15 16:16:00 [autocue2.metadata:3] ("liq_fade_in", "0.10")
2024/04/15 16:16:00 [autocue2.metadata:3] ("liq_fade_out", "0.10")
2024/04/15 16:16:00 [autocue2.metadata:3] ("liq_longtail", "false")
2024/04/15 16:16:00 [autocue2.metadata:3] ("liq_loudness", "-15.62 dB")
2024/04/15 16:16:00 [autocue2.metadata:3] ("replaygain_track_gain", "-2.39 dB")

[LordHelmchen666]

Simple solution. I added logo now when using your jingle script and don't use silence anymore. :)

[LordHelmchen666]

No good result. I need to find a solution to apply replaygain, but don't set fade and cue values. Any idea?

[LordHelmchen666]

I have added a short quiet sound at beginning and end. This is good enough for me now.

[LordHelmchen666]

Now i have found a good solution. I changed your jingle script to add a bed (background) instead a logo (at beginning and end). My random jingle now has every time same length (and loudness) and do not confuse autocue2 anymore. Sometimes it was cutoff, because i generate every minute new jingle.

Here are the scripts
jingle_bed

#!/bin/bash
# jingle
# 2023-05-30 Moonbase
# 2023-09-02 -- added gtts, simplified sox logic, adjustable sample rate
# 2023-10-23 -- use $(xdg-user-dir MUSIC) to find path to Music folder
#
# Create a MP3 Jingle with specified text.
# jingle "FILENAME" "Text to speak"
# FILENAME without extension (will be used as Title tag)
# Text to speak will be surrounded by logo sound, if specified
# and saved into the specified media folder
#
# Since TTS quality is usually rather low and most stations
# broadcast 128 kbit/s MP3 streams, it makes no sense blowing up the
# file size by artifically creating a "better quality". It is NOT.
# We use 2ch, 44.1kHz, 128kbit/s MP3 here.
#
# Needs: sox
# Optional: mid3iconv
# Engines: pico2wave, gtts-cli

me=`basename "$0"`

# Possible output languages for "pico" (pico2wave)
# (might require different sox post-processing)
# de-DE - German
# en-GB - British Englisch
# en-US - US Englisch
# fr-FR - French
# it-IT - Italian
# es-ES - Spanish

# "gtts" (gtts-cli) supports more languages, see 'gtts-cli --all'

lang="de-DE" # language for PicoTTS (Google TTS will use the first 2 letters)

# set to the TTS engine you prefer
# currently supported: pico (pico2wave), gtts (gtts-cli)
#engine="pico"
#engine="gtts"
engine="piper"

# **EDIT THIS!** Use a trailing slash for folder names ("/folder/").
# The $(xdg-user-dir MUSIC) points to the 'Music' folder in your home directory.
media_folder="/var/lib/docker/volumes/azuracast_station_data/_data/eisradio/media/Promos/"
# padding in seconds, at start, at end to the generated speech
pad="pad 4 0"
# Add a bed (background music) to your jingle
bed='/var/azuracast/sounds/pop-rock-logo-159921.mp3'
# station name (used as artist in ID3 tag)
station="Eisradio"
# some other tags you might want to preset
# Note: sox writes ID3v1.1 + ID3v2.3.0 tags
album="Promo"
genre="Jingles"

# In AzuraCast, edit the Jingle, and set, in the "Advanced" tab:
#   Custom Fading: Overlap Time (seconds): 0.1
#   Custom Fading: Fade-In Time (seconds): 0.1
#   Custom Fading: Fade-Out Time (seconds): 0.1 (0 ignores the fade-out)

# get current year for tag
YEAR=$(date +%Y)

# You shouldn’t need to change this. IF you do, remember that the logo file(s)
# MUST have the same sample rate!
samplerate=44100

show_help () {
echo "$me FILENAME \"TEXT TO SPEAK\""
echo "Generate MP3 Jingle in ${media_folder},"
echo "surrounded by logo sound."
echo ""
echo "Use FILENAME without extension (but with quoting if needed)."
echo "It will be used as the ID3 'title' tag."
exit 1
}

if [ "$1" = "" ] ; then
  show_help
fi

if [ "$2" = "" ] ; then
  show_help
fi

# function to check if required software is installed
#   is_installed COMMAND
#     returns 0 (installed) or 1 (not installed)
#   is_installed COMMAND value
#     returns 0 (installed) or exits with 'value' and displays an error message
is_installed() {
  if ! command -v "$1" > /dev/null ; then
    if [ ! -z "$2" ] ; then
      echo >&2 "$me: Error: '$1' not installed."
      exit $2
    else
      return 1
    fi
  fi
}

# check if sox installed, exit if not
is_installed sox 1

# build text
text="$2"

# use sox to make output better understandable (voices are rather muffled)
# adding some treble in the range of +3 to +6 dB helps
# some voices might need a little bass reduction, use s/th like "bass -6 400"
# to avoid clipping, give headroom (gain -h) and reclaim afterwards (gain -r)
# skip the "compand" part if you don’t want compression/limiting.

# sox effects to optimize TTS sound:
# these can vary depending on engine & language used (see CASE)
sox_effect=''

# sox compressor/limiter settings I use to optimize speech output
sox_compand='compand 0.008,0.024 -60,-60,-40,-35,0,-20 10 -60'

# create WAV audio file using PicoTTS
case $engine in

  pico)
    is_installed 'pico2wave' 1 # check if installed, exit if not
    #text='<volume level="60">'"$2</volume>"
    pico2wave -l=$lang -w=/tmp/jingle.wav "$text"
    sox_effect='bass +2 treble +3'
    sox -G /tmp/jingle.wav /tmp/jingle2.wav channels 2 rate $samplerate $sox_effect $pad
    ;;
  
  piper)
    is_installed 'piper' 1 # check if installed, exit if not
    echo "$text" |  piper --model /usr/local/bin/piper-tts-voices/de_DE-thorsten-high.onnx --output_file /tmp/jingle.wav
    sox_effect='bass +2 treble +1'
    sox -G /tmp/jingle.wav /tmp/jingle2.wav channels 2 rate $samplerate $sox_effect $pad
    ;;
    
  gtts)
    is_installed 'gtts-cli' 1 # check if installed, exit if not
    gtts-cli -l ${lang:0:2} -o /tmp/jingle.mp3 "$text"
    # Google TTS is a little slow, speed it up a little
    sox_effect='tempo 1.2'
    sox -G /tmp/jingle.mp3 /tmp/jingle2.wav channels 2 rate $samplerate $sox_effect $pad
    ;;
  
  *)
    echo >&2 "$me: Unknown TTS engine '$engine'"
    exit 1

esac

# SOX processing: Add bed, tags, effects, apply compression/limiting, pad.
if [ "$bed" != "" ] ; then
  # ensure # channels and sample rate match
  sox -G -v 0.05 "$bed" /tmp/bed.wav channels 2 rate $samplerate
  # combine and process
  sox -G -m /tmp/bed.wav /tmp/jingle2.wav \
    -c2 -r $samplerate -C 128 \
    --add-comment "Title=$1" \
    --add-comment "Artist=$station" \
    --add-comment "Album=$album" \
    --add-comment "Year=$YEAR" \
    --add-comment "Genre=$genre" \
    /tmp/jingle.mp3 \
    $sox_compand
else
  # make <jingle> MP3 only
  sox -G /tmp/jingle2.wav \
    -c2 -r $samplerate -C 128 \
    --add-comment "Title=$1" \
    --add-comment "Artist=$station" \
    --add-comment "Album=$album" \
    --add-comment "Year=$YEAR" \
    --add-comment "Genre=$genre" \
    /tmp/jingle.mp3 \
    $sox_compand \
    $pad
fi

# Use 'mid3iconv' from the Mutagen package (if available), to
# - repair language encoding
# - remove ID3v1 tag
# - convert to ID3v2.4
if is_installed mid3iconv ; then
  mid3iconv -q --remove-v1 /tmp/jingle.mp3
fi

# If loudgain is installed, also do ReplayGain calculation
# -I3 forces ID3v2.3, -I4 forces ID3v2.4 (default)
# -L forces lowercase replaygain tags (default uppercase)
if is_installed loudgain ; then
  loudgain -I3 -S -L -a -k -s e /tmp/jingle.mp3
fi

# copy it into our media folder
cp /tmp/jingle.mp3 "${media_folder}$1.mp3"
# diag only: play it loud (sox play replaygain seems not to work)
#play --replay-gain track "${media_folder}$1.mp3"

I ran randomjingle via cron every minute
randomjingle

#!/bin/bash
d=$(date +%u)  # day of week (1..7); 1 is Monday 
h=$(date +%H)  # hour (00..23)
m=$(date +%M)  # minute (00..59)
day=$(LC_TIME="de_DE.UTF-8" date +%A)  # Full Name of day e.g. Sunday
month=$(LC_TIME="de_DE.UTF-8" date +%B)  # Full Name of month e.g. February
dom=$(date +%d)  # day of month (e.g, 01 or 23) 

## Replace some numbers for correct spelling
# Hours
h="${h/00/Null}"
h="${h/01/Ein}"
h="${h/02/Zwei}"
h="${h/03/Drei}"
h="${h/04/Vier}"
h="${h/05/Fünf}"
h="${h/06/Sechs}"
h="${h/07/Sieben}"
h="${h/08/Acht}"
h="${h/09/Neun}"
# Minutes
m="${m/00/''}"
m="${m/01/Eins}"
m="${m/02/Zwei}"
m="${m/03/Drei}"
m="${m/04/Vier}"
m="${m/05/Fünf}"
m="${m/06/Sechs}"
m="${m/07/Sieben}"
m="${m/08/Acht}"
m="${m/09/Neun}"
# day of month
dom="${dom/01/Erste}"
dom="${dom/02/Zweite}"
dom="${dom/03/Dritte}"
dom="${dom/04/Vierte}"
dom="${dom/05/Fünfte}"
dom="${dom/06/Sechste}"
dom="${dom/07/Siebte}"
dom="${dom/08/Achte}"
dom="${dom/09/Neunte}"
dom="${dom/10/Zehnte}"
dom="${dom/11/Elfte}"
dom="${dom/12/Zwölfte}"
dom="${dom/13/Dreizehnte}"
dom="${dom/14/Vierzehnte}"
dom="${dom/15/Fünfzehnte}"
dom="${dom/16/Sechszehnte}"
dom="${dom/17/Siebzehnte}"
dom="${dom/18/Achtzehnte}"
dom="${dom/19/Neunzehnte}"
dom="${dom/20/Zwanzigste}"
dom="${dom/21/Einundzwangzigste}"
dom="${dom/22/Zweiundzwandzigste}"
dom="${dom/23/Dreiundzwandzigste}"
dom="${dom/24/Vierundzwandzigste}"
dom="${dom/25/Fünfundzwandzigste}"
dom="${dom/26/Sechsundzwandzigste}"
dom="${dom/27/Siebenundzwandzigste}"
dom="${dom/28/Achtundzwandzigste}"
dom="${dom/29/Neunundzwandzigste}"
dom="${dom/30/Dreissigste}"
dom="${dom/31/Einunddreissigste}"


array[0]="Sie hören Eis Radio. Mit uns durch den $day"
array[1]="Mit uns durch den $month. Wir wünschen weiterhin gute Unterhaltung bei Eis Radio."
array[2]="Es ist $h Uhr $m. Weiterhin viel Spass bei Eis Radio"
array[3]="Sie hören Eis Radio. Es ist $day $h Uhr $m"
array[4]="Heute ist $day der $dom $month. Es ist $h Uhr $m. Gute Unterhaltung mit Eis Radio."
array[5]="Sie hören Eis Radio. Es ist $day der $dom $month $h Uhr $m"
array[6]="Die beste Musik bei Eis Radio. Es ist $day der $dom $month $h Uhr $m"
array[7]="Die beste Musik bei Eis Radio. Es ist $h Uhr $m. Weiterhin gute Unterhaltung"
array[8]="Es ist $h Uhr $m. Mit uns durch den $day"

size=${#array[@]}
index=$(($RANDOM % $size))
text=$(echo ${array[$index]})
jingle_bed 'randomjingle' "$text"

## Use bash for loop // create all jingles at once for testing
#for index in ${!array[@]}; do jingle_bed 'randomjingle'$((index+1)) "${array[index]}" ; done

And here is an example
randomjingle.webm

[Moonbase59]

Hey @LordHelmchen666, sorry I totally missed this. For some reason I got no notification.

Great you found a solution and could use one of my scripts!

Some more ideas for your scripting:

• loudgain is automatically invoked if installed, so you should get ReplayGain data.
• Using autocue will also provide these, if "unify" is set.
• You could determine the duration in the jingle script, and have it set liq_cue_in, liq_cue_out and even liq_fade_in and liq_fade_out tags. These should overrule the automated calculations.
• You can, in Liquidsoap, also annotate these values (duration is in metadata "duration"), which would in turn override tags.

And of course this will all work with the new integrated autocue.cue_file, too.

[LordHelmchen666]

Thanks for your suggestions! But set liq_cue_out was the problem in my special case. I couldn't set manually because the jingle always had another length. That's why i underlay now with a music bed. Now the jingle always has 15 seconds length and i could set the values in Azuracast.
What happened before (without music bed)?
the jingle is generated every minute with another text. So one time it has 6 seconds another time it would have 10 seconds length. cue_file correctly determines 6 seconds and passes values to LS. Until playout the jingle is newly created with 10 second length. On playout LS correctly cue_out after 6 seconds and cut off the 10 second jingle (4 seconds to early).

So everything is working correctly!

[Moonbase59]

Yup, I have the same with my time announcement—length varies.
Glad that it works well for you!