Init commit

backend.py

@@ -0,0 +1,50 @@
+from flask import Flask, render_template
+from flask_socketio import SocketIO, emit
+# from flask_cors import CORS
+from audiocraft.models import musicgen
+import torchaudio
+import soundfile as sf
+
+
+# app = Flask(__name__)
+app = Flask(__name__, static_folder="../build", static_url_path="/")
+app.debug = True
+app.secret_key = "random secret key!"
+# CORS(app)
+# cors = CORS(app, resource={r"/*": {"origins": "*"}})
+socketio = SocketIO(app, cors_allowed_origins="*")
+print("Loading model...")
+model = musicgen.MusicGen.get_pretrained("melody")
+model.set_generation_params(duration=8)
+
+@app.route("/")
+def index():
+    print("HI")
+    return render_template("index.html")
+
+
+@socketio.on("connect")
+def connect():
+    print("Client connected")
+    stream_audio()
+
+
+@socketio.on("disconnect")
+def disconnect():
+    print("Client disconnected")
+
+def stream_audio(data):
+    descriptions = ["Film score epic moment"]
+    melody, sr = torchaudio.load("./asitwas_vocals.wav")
+    print("Running inference...")
+    wav = model.generate_with_chroma(descriptions, melody[None].expand(1, -1, -1), sr)
+    model_sampling_rate = 32000
+    sf.write("output.wav", wav[0].numpy().T, model_sampling_rate)
+    chunk_size = 1024
+    for i in range(0, len(wav[0]), chunk_size):
+        chunk = wav[0][i : i + chunk_size]# * 500
+        emit("audio_chunk", chunk.tolist())
+
+
+if __name__ == "__main__":
+    socketio.run(app)

inference.py

@@ -0,0 +1,16 @@
+from audiocraft.models import musicgen
+import torchaudio
+import soundfile as sf
+
+print("Loading model...")
+# model = musicgen.MusicGen.get_pretrained("melody")
+model = musicgen.MusicGen.get_pretrained("small")
+model.set_generation_params(duration=8)
+
+descriptions = ["Film score epic moment"]
+melody, sr = torchaudio.load("./asitwas_vocals.wav")
+print("Running inference...")
+# wav = model.generate_with_chroma(descriptions, melody[None].expand(1, -1, -1), sr)
+wav = model.generate(descriptions)  # generates 3 samples.
+model_sampling_rate = 32000
+sf.write("output.wav", wav[0].numpy().T, model_sampling_rate)

setup.py

@@ -0,0 +1,40 @@
+from pathlib import Path
+from setuptools import setup, find_packages
+
+NAME = "soundsauce"
+DESCRIPTION = ""
+URL = ""
+EMAIL = ""
+AUTHOR = ""
+REQUIRES_PYTHON = ">=3.9.0"
+VERSION = "0.0.1"
+
+HERE = Path(__file__).parent
+
+try:
+    with open(HERE / "README.md", encoding="utf-8") as f:
+        long_description = "\n" + f.read()
+except FileNotFoundError:
+    long_description = DESCRIPTION
+
+setup(
+    name=NAME,
+    version=VERSION,
+    description=DESCRIPTION,
+    long_description=long_description,
+    long_description_content_type="text/markdown",
+    author=AUTHOR,
+    author_email=EMAIL,
+    python_requires=REQUIRES_PYTHON,
+    url=URL,
+    packages=find_packages(),
+    install_requires=[
+        "torch>=2.0",
+        "torchaudio>=2.0.1",
+        "soundfile",
+        "flask",
+        "flask-socketio",
+        "audiocraft@git+https://github.com/facebookresearch/audiocraft"
+    ],
+    include_package_data=True,
+)