File size: 4,403 Bytes
c5c143f
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
from __future__ import annotations

import gradio as gr
import soundfile as sf

try:
    import spaces
except ImportError:
    class spaces:
        class GPU:
            def __init__(self, func=None, duration=180):
                self.func = func

            def __call__(self, *args, **kwargs):
                if self.func is not None:
                    return self.func(*args, **kwargs)
                return args[0]

from pyharp import ModelCard, build_endpoint

from stable_audio_runtime import edit_audio

MIN_AUDIO_SECONDS = 5
MAX_AUDIO_SECONDS = 30

model_card = ModelCard(
    name="Stable Audio 3",
    description=(
        "Restyle, inpaint, or continue music audio using a text prompt."
    ),
    author="Stability AI",
    tags=[
        "audio-generation",
        "music-editing",
        "audio-inpainting",
        "audio-continuation",
    ],
)


def _audio_duration(path: str | None) -> float:
    if not path:
        raise gr.Error("Please upload a music clip.")
    try:
        duration = float(sf.info(path).duration)
    except Exception as exc:
        raise gr.Error(f"Could not read the audio file: {exc}") from exc
    if duration < MIN_AUDIO_SECONDS:
        raise gr.Error(f"Audio must be at least {MIN_AUDIO_SECONDS} seconds.")
    if duration > MAX_AUDIO_SECONDS:
        raise gr.Error(f"Audio must be no longer than {MAX_AUDIO_SECONDS} seconds.")
    return duration


@spaces.GPU(duration=180)
def process_fn(
    audio_path: str | None,
    prompt: str,
    mode: str,
    edit_start: float,
    edit_end: float,
    continuation_length: float,
    strength: float,
    seed: int,
) -> str:
    duration = _audio_duration(audio_path)
    prompt = (prompt or "").strip()
    if not prompt:
        raise gr.Error("Please enter an editing prompt.")

    if mode == "Inpaint" and (
        edit_start < 0 or edit_end > duration or edit_end <= edit_start
    ):
        raise gr.Error(
            "The inpaint region must be inside the uploaded clip, "
            "with the end after the start."
        )

    try:
        return edit_audio(
            audio_path=audio_path,
            prompt=prompt,
            mode=mode,
            edit_start=float(edit_start),
            edit_end=float(edit_end),
            continuation_length=float(continuation_length),
            strength=float(strength),
            seed=int(seed),
        )
    except gr.Error:
        raise
    except Exception as exc:
        raise gr.Error(f"Stable Audio 3 inference failed: {exc}") from exc


with gr.Blocks(title="Stable Audio 3") as demo:
    input_components = [
        gr.Audio(type="filepath", label="Music Audio")
        .harp_required(True)
        .set_info("Music clip between 5 and 30 seconds."),
        gr.Textbox(
            label="Prompt",
            placeholder="A warm synthwave groove with punchy drums",
        ).harp_required(True),
        gr.Dropdown(
            choices=["Restyle", "Inpaint", "Continue"],
            value="Restyle",
            label="Edit Mode",
        ),
        gr.Slider(
            minimum=0,
            maximum=30,
            value=4,
            step=0.1,
            label="Edit Start (seconds)",
        ).set_info("Used by Inpaint mode."),
        gr.Slider(
            minimum=0,
            maximum=30,
            value=8,
            step=0.1,
            label="Edit End (seconds)",
        ).set_info("Used by Inpaint mode."),
        gr.Slider(
            minimum=1,
            maximum=15,
            value=8,
            step=1,
            label="Continuation Length (seconds)",
        ).set_info("Used by Continue mode."),
        gr.Slider(
            minimum=0.1,
            maximum=1.0,
            value=0.75,
            step=0.05,
            label="Transformation Strength",
        ).set_info("Used by Restyle mode."),
        gr.Slider(
            minimum=0,
            maximum=99999,
            value=0,
            step=1,
            label="Seed",
        ).set_info("Use 0 for a random seed."),
    ]
    output_components = [
        gr.Audio(type="filepath", label="Edited Audio"),
    ]
    build_endpoint(
        model_card=model_card,
        input_components=input_components,
        output_components=output_components,
        process_fn=process_fn,
    )


if __name__ == "__main__":
    demo.queue(default_concurrency_limit=1).launch(show_error=True, pwa=True)