File size: 5,972 Bytes
1dc20d9
4c10907
7d35d1e
a7334d4
 
6394bbc
deaa9a6
7d35d1e
a7334d4
1dc20d9
7d35d1e
 
 
 
 
eabc43b
6394bbc
deaa9a6
6394bbc
eabc43b
a7334d4
7d35d1e
d2cd103
eabc43b
a7334d4
 
df6c3f5
a7334d4
 
 
 
7d35d1e
a7334d4
 
 
 
 
 
 
 
 
 
 
 
7d35d1e
a7334d4
 
 
 
 
df6c3f5
a7334d4
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
30f9d01
a1ddd2f
 
 
 
 
30f9d01
7d35d1e
 
 
 
a7334d4
 
 
 
 
 
 
 
 
7d35d1e
 
 
a7334d4
 
 
 
a1ddd2f
a7334d4
a1ddd2f
7d35d1e
 
a1ddd2f
7d35d1e
 
 
 
 
 
 
 
 
 
 
a7334d4
7d35d1e
 
 
a7334d4
 
 
 
 
 
 
 
7d35d1e
a7334d4
 
 
 
 
 
 
7d35d1e
deaa9a6
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
import spaces
import gradio as gr
import os
import json
import tempfile
from pathlib import Path

from model import Jamify
from utils import json_to_text, text_to_json

# Initialize the Jamify model once
print("Initializing Jamify model...")
jamify_model = Jamify()
print("Jamify model ready.")



gr.set_static_paths(paths=[Path.cwd().absolute()])

@spaces.GPU(duration=100)
def generate_song(reference_audio, lyrics_text, style_prompt, duration):
    # We need to save the uploaded files to temporary paths to pass to the model
    reference_audio = reference_audio not in ("", None) and reference_audio or None
    
    # Convert text format to JSON and save to temporary file
    lyrics_json = text_to_json(lyrics_text)
    
    # Create temporary file for lyrics JSON
    with tempfile.NamedTemporaryFile(mode='w', suffix='.json', delete=False) as f:
        json.dump(lyrics_json, f, indent=2)
        lyrics_file = f.name

    try:
        output_path = jamify_model.predict(
            reference_audio_path=reference_audio,
            lyrics_json_path=lyrics_file,
            style_prompt=style_prompt,
            duration_sec=duration
        )
        return output_path
    finally:
        # Clean up temporary file
        if os.path.exists(lyrics_file):
            os.unlink(lyrics_file)

# Load and cache examples
def load_examples():
    """Load examples from the examples directory and pre-compute text formats"""
    examples = []
    examples_file = "examples/input.json"
    
    if os.path.exists(examples_file):
        print("Loading and caching examples...")
        with open(examples_file, 'r') as f:
            examples_data = json.load(f)
            
        for example in examples_data:
            example_id = example.get('id', '')
            audio_path = example.get('audio_path', '')
            lrc_path = example.get('lrc_path', '')
            duration = example.get('duration', 120)
            
            # Load lyrics and convert to text format (pre-computed/cached)
            lyrics_text = ""
            if os.path.exists(lrc_path):
                try:
                    with open(lrc_path, 'r') as f:
                        lyrics_json = json.load(f)
                    lyrics_text = json_to_text(lyrics_json)
                    print(f"Cached example {example_id}: {len(lyrics_text)} chars")
                except Exception as e:
                    print(f"Error loading lyrics from {lrc_path}: {e}")
            
            examples.append({
                'id': example_id,
                'audio_path': audio_path if os.path.exists(audio_path) else None,
                'lyrics_text': lyrics_text,
                'duration': duration
            })
    
    print(f"Loaded {len(examples)} cached examples")
    return examples

def load_example(example_idx, examples):
    """Load a specific example and return its data"""
    if 0 <= example_idx < len(examples):
        example = examples[example_idx]
        return (
            example['audio_path'],
            example['lyrics_text'],
            example['duration']
        )
    return None, "", 120

# Load examples at startup
examples = load_examples()

# Get default values from first example
default_audio = examples[0]['audio_path'] if examples else None
default_lyrics = examples[0]['lyrics_text'] if examples else ""
default_duration = examples[0]['duration'] if examples else 120

# Gradio interface
with gr.Blocks() as demo:
    gr.Markdown("# Jamify: Music Generation from Lyrics and Style")
    gr.Markdown("Provide your lyrics, a style reference (either an audio file or a text prompt), and a desired duration to generate a song.")
    
    # Sample buttons section
    if examples:
        gr.Markdown("### Sample Examples")
        with gr.Row():
            example_buttons = []
            for i, example in enumerate(examples):
                button = gr.Button(f"Example {example['id']}", variant="secondary", size="sm")
                example_buttons.append(button)
    
    with gr.Row():
        with gr.Column():
            gr.Markdown("### Inputs")
            lyrics_text = gr.Textbox(
                label="Lyrics", 
                lines=10, 
                placeholder="Enter lyrics in format: word[start:end] word[start:end]...\nExample: It's[4.96:5.52] a[5.52:5.84] long[5.84:6.16] way[6.16:6.48]...",
                value=default_lyrics
            )
            duration_slider = gr.Slider(minimum=5, maximum=230, value=default_duration, step=30, label="Duration (seconds)")
            
            with gr.Tab("Style from Audio"):
                reference_audio = gr.File(label="Reference Audio (.mp3, .wav)", type="filepath", value=default_audio)
            with gr.Tab("Style from Text"):
                style_prompt = gr.Textbox(label="Style Prompt", lines=3, placeholder="e.g., A high-energy electronic dance track with a strong bassline and euphoric synths.")
            
            generate_button = gr.Button("Generate Song", variant="primary")
            
        with gr.Column():
            gr.Markdown("### Output")
            output_audio = gr.Audio(label="Generated Song")

    generate_button.click(
        fn=generate_song,
        inputs=[reference_audio, lyrics_text, style_prompt, duration_slider],
        outputs=output_audio,
        api_name="generate_song"
    )
    
    # Connect example buttons to load data
    if examples:
        for i, button in enumerate(example_buttons):
            button.click(
                fn=lambda idx=i: load_example(idx, examples),
                outputs=[reference_audio, lyrics_text, duration_slider]
            )

# Create necessary temporary directories for Gradio
print("Creating temporary directories...")
try:
    os.makedirs("/tmp/gradio", exist_ok=True)
    print("Temporary directories created successfully.")
except Exception as e:
    print(f"Warning: Could not create temporary directories: {e}")

demo.queue().launch()