metadata

license: mit
language: en
base_model: facebook/musicgen-medium
pipeline_tag: text-to-audio
tags:
  - audio
  - music-generation
  - text-to-music
  - musicgen
  - transformers
library_name: transformers
metrics:
  - type: audio_quality
    value: '32000'
    name: sample_rate
  - type: generation_length
    value: '30'
    name: max_duration_seconds
datasets: []

MelodyMaster V1

MelodyMaster V1 is an AI music generation model based on Meta's MusicGen-medium architecture. It generates high-quality music from text descriptions.

Model Description

Model Architecture: MusicGen (3.3B parameters)
Base Model: facebook/musicgen-medium
Task: Text-to-Music Generation
Output: 32kHz audio samples
Max Duration: 30 seconds

Usage

from transformers import AutoProcessor, MusicgenForConditionalGeneration

# Load model and processor
model = MusicgenForConditionalGeneration.from_pretrained("opentunesai/melodymasterv1")
processor = AutoProcessor.from_pretrained("opentunesai/melodymasterv1")

# Process text and generate music
inputs = processor(
    text=["happy rock song with electric guitar"],
    padding=True,
    return_tensors="pt",
)

audio_values = model.generate(**inputs, max_new_tokens=1500)

Example Prompts

"An upbeat electronic dance track with a strong beat"
"A peaceful piano melody with soft strings"
"A rock song with electric guitar and drums"
"Jazz trio with piano, bass and drums"

Demo

Try the model in our Gradio Demo Space

License

Apache 2.0

Acknowledgments

Based on Meta's MusicGen model. Original model card: facebook/musicgen-medium

Citation

@article{copet2023simple,
      title={Simple and Controllable Music Generation}, 
      author={Jade Copet and Felix Kreuk and Itai Gat and Tal Remez and David Kant and Gabriel Synnaeve and Yossi Adi and Alexandre Défossez},
      year={2023},
      journal={arXiv preprint arXiv:2306.05284},
}