import gradio as gr import os import shutil import spaces import sys # we will clone the repo and install the dependencies # NOTE: Still fixing bugs, not release, do not try :) ! os.system('pip install -r qa_mdt/requirements.txt') os.system('pip install xformers==0.0.26.post1') os.system('pip install torchlibrosa==0.0.9 librosa==0.9.2') os.system('pip install -q pytorch_lightning==2.1.3 torchlibrosa==0.0.9 librosa==0.9.2 ftfy==6.1.1 braceexpand') os.system('pip install torch==2.3.0+cu121 torchvision==0.18.0+cu121 torchaudio==2.3.0 --index-url https://download.pytorch.org/whl/cu121') # only then import the necessary modules from qa_mdt from qa_mdt.pipeline import MOSDiffusionPipeline pipe = MOSDiffusionPipeline() # this runs the pipeline with user input and saves the output as 'awesome.wav' @spaces.GPU() def generate_waveform(description): pipe(description) generated_file_path = "./awesome.wav" if os.path.exists(generated_file_path): return generated_file_path else: return "Error: Failed to generate the waveform." intro = """ # ๐ŸŽถ OpenMusic: AI-Powered Music Diffusion ๐ŸŽถ ![OpenMusic Banner](./banner.png) Welcome to **OpenMusic**, a next-gen diffusion model designed to generate high-quality audio from text descriptions! Simply enter a description of the music you'd like to hear, and our AI will generate it for you. --- ### Powered by: - [GitHub Repository](https://github.com/ivcylc/qa-mdt) by [@changli](https://github.com/ivcylc) ๐ŸŽ“. - Introduced in this [Paper](https://arxiv.org/pdf/2405.15863) - Hugging Face Diffusers Implementation ๐Ÿงจ (Super easy to use): [Model](https://huggingface.co/jadechoghari/qa_mdt) by [@jadechoghari](https://github.com/jadechoghari) ๐Ÿค—. --- """ # gradio interface iface = gr.Interface( description=intro, fn=generate_waveform, inputs=gr.Textbox(lines=2, placeholder="Enter a music description here..."), outputs=gr.Audio(label="Download the Music ๐ŸŽผ"), description="Enter a music description, and the model will generate a corresponding audio waveform. Download the output as 'awesome.wav'.", examples=[ ["A modern synthesizer creating futuristic soundscapes."], ["Acoustic ballad with heartfelt lyrics and soft piano."] ], cache_examples=True ) # Launch the Gradio app if __name__ == "__main__": iface.launch()