app.py

import gradio as gr
import librosa
import numpy as np
import torch

from diffusers import SpectrogramDiffusionPipeline, MidiProcessor

pipe = SpectrogramDiffusionPipeline.from_pretrained("google/music-spectrogram-diffusion", torch_dtype=torch.float16).to("cuda")
pipe.enable_xformers_memory_efficient_attention()

processor = MidiProcessor()


def predict(audio_file_pth):

    with torch.inference_mode():
        output = pipe(processor(audio_file_pth.name)[:2])
        audio = output.audios[0]

    return (16000, audio.ravel())


title = "Music Spectrogram Diffusion: Multi-instrument Music Synthesis with Spectrogram Diffusion"

description = """
In this work, the authors focus on a middle ground of neural synthesizers that can generate audio from MIDI sequences with arbitrary combinations of instruments in realtime. 
This enables training on a wide range of transcription datasets with a single model, which in turn offers note-level control of composition and instrumentation across a wide range of instruments.

They use a simple two-stage process: MIDI to spectrograms with an encoder-decoder Transformer, then spectrograms to audio with a generative adversarial network (GAN) spectrogram inverter.
"""
examples = []


gr.Interface(
    fn=predict,
    inputs=[
        gr.File(label="Upload MIDI", file_count="single", file_types=[".mid"]),
    ],
    outputs=[
        gr.Audio(label="Synthesised Music", type="numpy"),
    ],
    title=title,
    description=description,
    theme='gstaff/xkcd',
).launch(debug=True)