app.py

from audiocraft.models import MusicGen
import streamlit as st
import torch
import torchaudio
import io
import base64

@st.cache_resource
def load_model():
    model = MusicGen.get_pretrained('facebook/musicgen-small')
    return model

def generate_music_tensors(description, duration: int):
    st.write(f"Generating music for: '{description}' (Duration: {duration}s)")
    model = load_model()
    model.set_generation_params(use_sampling=True, top_k=250, duration=duration)
    output = model.generate(descriptions=[description], progress=True)
    return output[0]

def create_audio_buffer(samples: torch.Tensor):
    """Generate an in-memory audio buffer."""
    sample_rate = 32000
    samples = samples.detach().cpu()

    if samples.dim() == 2:
        samples = samples[None, ...]

    # Create an in-memory buffer to store the audio
    buffer = io.BytesIO()
    torchaudio.save(buffer, samples[0], sample_rate, format="wav")
    buffer.seek(0)
    return buffer

def generate_download_link(buffer, file_label="Download Music"):
    """Create a download link for the generated audio."""
    data = buffer.read()
    b64 = base64.b64encode(data).decode()
    href = f'<a href="data:audio/wav;base64,{b64}" download="generated_music.wav">{file_label}</a>'
    return href

# Apply CSS for improved UI styling
st.markdown(
    """
    <style>
    .title {
        font-size: 3em;
        text-align: center;
        color: #4A90E2;
        margin-top: 0;
    }
    .footer {
        position: fixed;
        left: 0;
        bottom: 0;
        width: 100%;
        background-color: #f1f1f1;
        text-align: center;
        padding: 10px;
        font-size: 0.8em;
        color: #555;
    }
    </style>
    """, unsafe_allow_html=True
)

# Set Streamlit page configuration

def main():
    st.markdown('<h1 class="title">Theaimart: Music Generator 🎵</h1>', unsafe_allow_html=True)
    st.write("Generate music based on your text input using Meta's Audiocraft library!")

    description = st.text_area("Enter a description:")
    duration = st.slider("Select duration (seconds)", 1, 20, 10)

    if description and duration:
        music_tensors = generate_music_tensors(description, duration)
        audio_buffer = create_audio_buffer(music_tensors)

        st.audio(audio_buffer, format="audio/wav")
        st.markdown(generate_download_link(audio_buffer), unsafe_allow_html=True)

    # Add footer message
    st.markdown('<div class="footer">Made with ❤️ by Theaimart</div>', unsafe_allow_html=True)

if __name__ == "__main__":
    main()