Add a server for the Singaporean Voice

server/Dockerfile

@@ -0,0 +1,15 @@
+# 1 
+FROM python:3.7
+
+# 2
+RUN pip install numpy Flask gunicorn malaya malaya-speech pydub
+
+# 3
+COPY src/ /app
+WORKDIR /app
+
+# 4
+ENV PORT 8080
+
+# 5
+CMD exec gunicorn --bind :$PORT --workers 1 --threads 8 app:app

server/src/app.py

@@ -0,0 +1,25 @@
+from flask import Flask, jsonify, send_file, request, abort
+import os
+import time
+import malaya_speech
+from pydub import AudioSegment
+from io import BytesIO
+import numpy as np
+
+
+app = Flask(__name__)
+
+
+@app.route('/', methods=['POST'])
+def speakSinglish(text):
+    if not request.json or not 'text' in request.json: abort(400)
+    mp3 = BytesIO()
+    AudioSegment((np.iinfo(np.int16).max * vocoder(tacotron.predict(request.json["text"])["universal-output"])).astype(np.int16).tobytes(), frame_rate=22050, sample_width=2, channels=1).export(mp3, format="mp3", bitrate="320k")
+    mp3.seek(0)
+    return send_file(mp3, as_attachment=True, attachment_filename='test.mp3', mimetype='audio/mpeg')
+
+
+if __name__ == '__main__':
+    tacotron = malaya_speech.tts.tacotron2(model='female-singlish')
+    vocoder = malaya_speech.vocoder.melgan(model='universal-1024')
+    app.run(debug=True, host='0.0.0.0', port=int(os.environ.get('PORT', 8080)))

test-server/test-malaya.py

@@ -0,0 +1,14 @@
+import numpy as np
+import matplotlib.pyplot as plt
+from pydub import AudioSegment
+from io import BytesIO
+import malaya_speech
+
+tacotron = malaya_speech.tts.tacotron2(model='female-singlish')
+vocoder = malaya_speech.vocoder.melgan(model='universal-1024')
+
+mp3 = BytesIO()
+AudioSegment((np.iinfo(np.int16).max * vocoder(tacotron.predict("Hello World, this is a test of a Singaporean accent. Currently we only have a female singlish accent but let's see how it goes.")["universal-output"])).astype(np.int16).tobytes(), frame_rate=22050, sample_width=2, channels=1).export(mp3, format="mp3", bitrate="320k")
+mp3.seek(0)
+
+print(mp3.read())