WIP

somleng · Oct 20, 2023 · 2cf5649 · 2cf5649
1 parent df950a3
commit 2cf5649
Show file tree

Hide file tree

Showing 3 changed files with 11 additions and 6 deletions.
diff --git a/components/app/Gemfile.lock b/components/app/Gemfile.lock
@@ -49,7 +49,7 @@ GIT
 
 GIT
   remote: https://github.com/somleng/tts_voices.git
-  revision: e719a88769a77d08a85dc3aa99ef04ba4630ad18
+  revision: 65a7b3024a27fb4c56dc03a86eb286fa1d83c24f
   specs:
     tts_voices (0.1.0)
       aws-sdk-polly
@@ -72,7 +72,7 @@ GEM
     adhearsion-loquacious (1.9.3)
     ast (2.4.2)
     aws-eventstream (1.2.0)
-    aws-partitions (1.835.0)
+    aws-partitions (1.838.0)
     aws-sdk-core (3.185.1)
       aws-eventstream (~> 1, >= 1.0.2)
       aws-partitions (~> 1, >= 1.651.0)

diff --git a/components/freeswitch/bin/aws_polly b/components/freeswitch/bin/aws_polly
@@ -3,12 +3,13 @@
 text="$1"
 file="$2"
 voice_id="$3"
-cache_file="$4"
+engine="$4"
+cache_file="$5"
 
 # Output the text to a file before executing polly
 # https://github.com/somleng/somleng-switch/pull/238
 echo "$text" > "$file.txt"
-aws polly synthesize-speech --sample-rate "8000" --output-format mp3 --voice-id "$voice_id" --text "file://$file.txt" "$file.mp3"
+aws polly synthesize-speech --sample-rate "8000" --output-format mp3 --voice-id "$voice_id" --engine "$engine" --text "file://$file.txt" "$file.mp3"
 
 ffmpeg -i "$file.mp3" "$cache_file"
 ln -s "$cache_file" "$file"

diff --git a/components/freeswitch/bin/cloud_tts b/components/freeswitch/bin/cloud_tts
@@ -6,6 +6,10 @@ voice="$3"
 cache_file="$4"
 
 if [[ "$voice" =~ ^Polly\..+ ]]; then
-  # extract Polly.from voice and execute aws_polly
-  aws_polly "$text" "$file" "${voice#"Polly."}" "$cache_file"
+  # extract 'Polly.' from voice
+  voice_id_with_engine="${voice#"Polly."}"
+  # extract '-Neural' from voice
+  voice_id="${voice_id_with_engine%"-Neural"}"
+  [[ $voice_id_with_engine = $voice_id ]] && engine="standard" || engine="neural"
+  aws_polly "$text" "$file" "$voice_id" "$engine" "$cache_file"
 fi