M5Stack Module-LLMでVOICEVOXを使ってテキストから音声を再生するスクリプト

llm_voicevox.sh

#!/bin/bash
in_file=$1

## 設定
speaker=1    # ずんだもん
#speaker=2    # 四国めたん
volume=0.1

if [ ! -e "${in_file}" ] ; then
  echo "input file not found"
  exit 1
fi

echo -n "Generate audio query..."
curl -sGX POST \
  "http://127.0.0.1:50021/audio_query" \
  --data-urlencode "text@${in_file}" \
  --data "speaker=${speaker}" \
  --data "enable_katakana_english=true" \
  -H "Accept: application/json" \
  --output /tmp/vv_query.json
echo done
ls -l /tmp/vv_query.json

jq ".outputSamplingRate=16000 | .outputStereo=true | .volumeScale=${volume}" \
  /tmp/vv_query.json > /tmp/vv_query.json.tmp \
  && mv /tmp/vv_query.json.tmp /tmp/vv_query.json

echo -n "Synthesis..."
curl -sX POST \
  "http://127.0.0.1:50021/synthesis?speaker=${speaker}&enable_interrogative_upspeak=true" \
  -H "Accept: audio/wav" \
  -H "Content-Type: application/json" \
  --data @/tmp/vv_query.json \
  --output /tmp/vv_voice.wav
echo done
ls -l /tmp/vv_voice.wav

echo -n "Play..."
#aplay -D plughw:0,1 /tmp/vv_voice.wav
tinyplay -D0 -d1 /tmp/vv_voice.wav
echo done

rm -f /tmp/vv_query.json /tmp/vv_voice.wav