""" Copyright (c) Meta Platforms, Inc. and affiliates. All rights reserved. This source code is licensed under the license found in the LICENSE file in the root directory of this source tree. """ import argparse from concurrent.futures import ProcessPoolExecutor import subprocess as sp from tempfile import NamedTemporaryFile import time import warnings import torch import gradio as gr from audiocraft.data.audio_utils import convert_audio from audiocraft.data.audio import audio_write from audiocraft.models import MusicGen MODEL = None _old_call = sp.call def _call_nostderr(*args, **kwargs): # Avoid ffmpeg vomitting on the logs. kwargs['stderr'] = sp.DEVNULL kwargs['stdout'] = sp.DEVNULL _old_call(*args, **kwargs) sp.call = _call_nostderr pool = ProcessPoolExecutor(3) pool.__enter__() def make_waveform(*args, **kwargs): be = time.time() with warnings.catch_warnings(): warnings.simplefilter('ignore') out = gr.make_waveform(*args, **kwargs) print("Make a video took", time.time() - be) return out def load_model(): print("Loading model") return MusicGen.get_pretrained("melody") def predict(texts, melodies): global MODEL if MODEL is None: MODEL = load_model() duration = 12 max_text_length = 512 texts = [text[:max_text_length] for text in texts] MODEL.set_generation_params(duration=duration) print("new batch", len(texts), texts, [None if m is None else (m[0], m[1].shape) for m in melodies]) be = time.time() processed_melodies = [] target_sr = 32000 target_ac = 1 for melody in melodies: if melody is None: processed_melodies.append(None) else: sr, melody = melody[0], torch.from_numpy(melody[1]).to(MODEL.device).float().t() if melody.dim() == 1: melody = melody[None] melody = melody[..., :int(sr * duration)] melody = convert_audio(melody, sr, target_sr, target_ac) processed_melodies.append(melody) outputs = MODEL.generate_with_chroma( descriptions=texts, melody_wavs=processed_melodies, melody_sample_rate=target_sr, progress=False ) outputs = outputs.detach().cpu().float() out_files = [] for output in outputs: with NamedTemporaryFile("wb", suffix=".wav", delete=False) as file: audio_write( file.name, output, MODEL.sample_rate, strategy="loudness", loudness_headroom_db=16, loudness_compressor=True, add_suffix=False) out_files.append(pool.submit(make_waveform, file.name)) res = [[out_file.result() for out_file in out_files]] print("batch finished", len(texts), time.time() - be) return res def ui(**kwargs): with gr.Blocks() as demo: gr.Markdown( """ # MusicGen This is the demo for [MusicGen](https://github.com/facebookresearch/audiocraft), a simple and controllable model for music generation presented at: ["Simple and Controllable Music Generation"](https://huggingface.co/papers/2306.05284).
Duplicate Space for longer sequences, more control and no queue.

""" ) with gr.Row(): with gr.Column(): with gr.Row(): text = gr.Text(label="Describe your music", lines=2, interactive=True) melody = gr.Audio(source="upload", type="numpy", label="Condition on a melody (optional)", interactive=True) with gr.Row(): submit = gr.Button("Generate") with gr.Column(): output = gr.Video(label="Generated Music") submit.click(predict, inputs=[text, melody], outputs=[output], batch=True, max_batch_size=8) gr.Examples( fn=predict, examples=[ [ "An 80s driving pop song with heavy drums and synth pads in the background", "./assets/bach.mp3", ], [ "A cheerful country song with acoustic guitars", "./assets/bolero_ravel.mp3", ], [ "90s rock song with electric guitar and heavy drums", None, ], [ "a light and cheerly EDM track, with syncopated drums, aery pads, and strong emotions bpm: 130", "./assets/bach.mp3", ], [ "lofi slow bpm electro chill with organic samples", None, ], ], inputs=[text, melody], outputs=[output] ) gr.Markdown(""" ### More details The model will generate 12 seconds of audio based on the description you provided. You can optionaly provide a reference audio from which a broad melody will be extracted. The model will then try to follow both the description and melody provided. All samples are generated with the `melody` model. You can also use your own GPU or a Google Colab by following the instructions on our repo. See [github.com/facebookresearch/audiocraft](https://github.com/facebookresearch/audiocraft) for more details. """) # Show the interface launch_kwargs = {} username = kwargs.get('username') password = kwargs.get('password') server_port = kwargs.get('server_port', 0) inbrowser = kwargs.get('inbrowser', False) share = kwargs.get('share', False) server_name = kwargs.get('listen') launch_kwargs['server_name'] = server_name if username and password: launch_kwargs['auth'] = (username, password) if server_port > 0: launch_kwargs['server_port'] = server_port if inbrowser: launch_kwargs['inbrowser'] = inbrowser if share: launch_kwargs['share'] = share demo.queue(max_size=60).launch(**launch_kwargs) if __name__ == "__main__": parser = argparse.ArgumentParser() parser.add_argument( '--listen', type=str, default='127.0.0.1', help='IP to listen on for connections to Gradio', ) parser.add_argument( '--username', type=str, default='', help='Username for authentication' ) parser.add_argument( '--password', type=str, default='', help='Password for authentication' ) parser.add_argument( '--server_port', type=int, default=0, help='Port to run the server listener on', ) parser.add_argument( '--inbrowser', action='store_true', help='Open in browser' ) parser.add_argument( '--share', action='store_true', help='Share the gradio UI' ) args = parser.parse_args() ui( username=args.username, password=args.password, inbrowser=args.inbrowser, server_port=args.server_port, share=args.share, listen=args.listen )