Spaces:
Running
on
Zero
Running
on
Zero
Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,48 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import torch
|
2 |
+
import soundfile as sf
|
3 |
+
from diffusers import StableAudioPipeline
|
4 |
+
import gradio as gr
|
5 |
+
|
6 |
+
# Load the StableAudio pipeline model
|
7 |
+
pipe = StableAudioPipeline.from_pretrained("stabilityai/stable-audio-open-1.0", torch_dtype=torch.float16)
|
8 |
+
pipe = pipe.to("cuda")
|
9 |
+
|
10 |
+
# Define the function to generate the sound based on a text prompt
|
11 |
+
@spaces.GPU
|
12 |
+
def generate_sound(prompt, negative_prompt, seed, inference_steps, duration, waveforms):
|
13 |
+
# Set the seed for reproducibility
|
14 |
+
generator = torch.Generator("cuda").manual_seed(seed)
|
15 |
+
|
16 |
+
# Run the audio generation
|
17 |
+
audio = pipe(
|
18 |
+
prompt,
|
19 |
+
negative_prompt=negative_prompt,
|
20 |
+
num_inference_steps=inference_steps,
|
21 |
+
audio_end_in_s=duration,
|
22 |
+
num_waveforms_per_prompt=waveforms,
|
23 |
+
generator=generator,
|
24 |
+
).audios
|
25 |
+
|
26 |
+
# Get the output and save to a file
|
27 |
+
output = audio[0].T.float().cpu().numpy()
|
28 |
+
sf.write("generated_sound.wav", output, pipe.vae.sampling_rate)
|
29 |
+
|
30 |
+
return "generated_sound.wav"
|
31 |
+
|
32 |
+
# Define the Gradio interface
|
33 |
+
app = gr.Interface(
|
34 |
+
fn=generate_sound,
|
35 |
+
inputs=[
|
36 |
+
gr.inputs.Textbox(label="Text Prompt", placeholder="Describe the sound you'd like to generate..."),
|
37 |
+
gr.inputs.Textbox(label="Negative Prompt", placeholder="Describe what you don't want in the sound..."),
|
38 |
+
gr.inputs.Slider(label="Seed", minimum=0, maximum=10000, step=1, default=0),
|
39 |
+
gr.inputs.Slider(label="Inference Steps", minimum=50, maximum=500, step=10, default=200),
|
40 |
+
gr.inputs.Slider(label="Duration (seconds)", minimum=1.0, maximum=30.0, step=0.5, default=10.0),
|
41 |
+
gr.inputs.Slider(label="Number of Waveforms", minimum=1, maximum=5, step=1, default=1)
|
42 |
+
],
|
43 |
+
outputs=gr.Audio(label="Generated Sound"),
|
44 |
+
title="StableAudio Text-to-Speech Generator",
|
45 |
+
description="Generate high-quality audio from text using StableAudio."
|
46 |
+
)
|
47 |
+
|
48 |
+
app.launch()
|