Spaces:
Running
on
Zero
Running
on
Zero
Commit
•
0669a33
1
Parent(s):
6d3fdbe
make it multiplayer (#3)
Browse files- make it multiplayer (9f88b4455c91e04313c6be362fd8ac3713b4d5bf)
- Update README.md (a2fa3a420dbb1f91b30a5b8f7b4398333f2c6d89)
Co-authored-by: Apolinário from multimodal AI art <multimodalart@users.noreply.huggingface.co>
README.md
CHANGED
@@ -1,5 +1,5 @@
|
|
1 |
---
|
2 |
-
title:
|
3 |
emoji: 💻
|
4 |
colorFrom: pink
|
5 |
colorTo: red
|
|
|
1 |
---
|
2 |
+
title: Stable Audio Live Multiplayer
|
3 |
emoji: 💻
|
4 |
colorFrom: pink
|
5 |
colorTo: red
|
app.py
CHANGED
@@ -10,6 +10,13 @@ import uuid
|
|
10 |
from stable_audio_tools import get_pretrained_model
|
11 |
from stable_audio_tools.inference.generation import generate_diffusion_cond
|
12 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
13 |
# Load the model outside of the GPU-decorated function
|
14 |
def load_model():
|
15 |
|
@@ -19,7 +26,7 @@ def load_model():
|
|
19 |
|
20 |
# Function to set up, generate, and process the audio
|
21 |
@spaces.GPU(duration=120) # Allocate GPU only when this function is called
|
22 |
-
def generate_audio(prompt, sampler_type_dropdown, seconds_total=30, steps=100, cfg_scale=7,sigma_min_slider=0.3,sigma_max_slider=500):
|
23 |
print(f"Prompt received: {prompt}")
|
24 |
print(f"Settings: Duration={seconds_total}s, Steps={steps}, CFG Scale={cfg_scale}")
|
25 |
|
@@ -76,34 +83,54 @@ def generate_audio(prompt, sampler_type_dropdown, seconds_total=30, steps=100, c
|
|
76 |
print(f"Audio trimmed to {seconds_total} seconds.")
|
77 |
|
78 |
# Generate a unique filename for the output
|
79 |
-
|
|
|
|
|
80 |
print(f"Saving audio to file: {unique_filename}")
|
81 |
|
82 |
# Save to file
|
83 |
torchaudio.save(unique_filename, output, sample_rate)
|
84 |
print(f"Audio saved: {unique_filename}")
|
85 |
|
|
|
|
|
|
|
86 |
# Return the path to the generated audio file
|
87 |
return unique_filename
|
88 |
|
89 |
-
|
90 |
-
|
91 |
-
|
92 |
-
|
93 |
-
|
94 |
-
|
95 |
-
|
96 |
-
|
97 |
-
|
98 |
-
|
99 |
-
|
100 |
-
|
101 |
-
|
102 |
-
|
103 |
-
|
104 |
-
|
105 |
-
|
106 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
107 |
[
|
108 |
"Create a serene soundscape of a quiet beach at sunset.", # Text prompt
|
109 |
"dpmpp-2m-sde", # Sampler type
|
@@ -157,12 +184,70 @@ interface = gr.Interface(
|
|
157 |
0.3, # Sigma min
|
158 |
500 # Sigma max
|
159 |
]
|
160 |
-
|
161 |
-
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
162 |
|
163 |
-
# Pre-load the model to avoid multiprocessing issues
|
164 |
model, model_config = load_model()
|
165 |
|
166 |
-
|
167 |
-
interface.queue(max_size=10).launch()
|
168 |
-
|
|
|
10 |
from stable_audio_tools import get_pretrained_model
|
11 |
from stable_audio_tools.inference.generation import generate_diffusion_cond
|
12 |
|
13 |
+
PAGE_SIZE = 10
|
14 |
+
FILE_DIR_PATH = "/data"
|
15 |
+
|
16 |
+
theme = gr.themes.Base(
|
17 |
+
font=[gr.themes.GoogleFont('Libre Franklin'), gr.themes.GoogleFont('Public Sans'), 'system-ui', 'sans-serif'],
|
18 |
+
)
|
19 |
+
|
20 |
# Load the model outside of the GPU-decorated function
|
21 |
def load_model():
|
22 |
|
|
|
26 |
|
27 |
# Function to set up, generate, and process the audio
|
28 |
@spaces.GPU(duration=120) # Allocate GPU only when this function is called
|
29 |
+
def generate_audio(prompt, sampler_type_dropdown, seconds_total=30, steps=100, cfg_scale=7,sigma_min_slider=0.3,sigma_max_slider=500, progress=gr.Progress(track_tqdm=True)):
|
30 |
print(f"Prompt received: {prompt}")
|
31 |
print(f"Settings: Duration={seconds_total}s, Steps={steps}, CFG Scale={cfg_scale}")
|
32 |
|
|
|
83 |
print(f"Audio trimmed to {seconds_total} seconds.")
|
84 |
|
85 |
# Generate a unique filename for the output
|
86 |
+
random_uuid = uuid.uuid4().hex
|
87 |
+
unique_filename = f"/data/output_{random_uuid}.wav"
|
88 |
+
unique_textfile = f"/data/output_{random_uuid}.txt"
|
89 |
print(f"Saving audio to file: {unique_filename}")
|
90 |
|
91 |
# Save to file
|
92 |
torchaudio.save(unique_filename, output, sample_rate)
|
93 |
print(f"Audio saved: {unique_filename}")
|
94 |
|
95 |
+
with open(unique_textfile, "w") as file:
|
96 |
+
file.write(prompt)
|
97 |
+
|
98 |
# Return the path to the generated audio file
|
99 |
return unique_filename
|
100 |
|
101 |
+
def list_all_outputs(generation_history):
|
102 |
+
directory_path = FILE_DIR_PATH
|
103 |
+
files_in_directory = os.listdir(directory_path)
|
104 |
+
wav_files = [os.path.join(directory_path, file) for file in files_in_directory if file.endswith('.wav')]
|
105 |
+
wav_files.sort(key=lambda x: os.path.getmtime(os.path.join(directory_path, x)), reverse=True)
|
106 |
+
history_list = generation_history.split(',') if generation_history else []
|
107 |
+
updated_files = [file for file in wav_files if file not in history_list]
|
108 |
+
updated_history = updated_files + history_list
|
109 |
+
return ','.join(updated_history), gr.update(visible=True)
|
110 |
+
|
111 |
+
def increase_list_size(list_size):
|
112 |
+
return list_size+PAGE_SIZE
|
113 |
+
|
114 |
+
css = '''
|
115 |
+
#live_gen:before {
|
116 |
+
content: '';
|
117 |
+
animation: svelte-z7cif2-pulseStart 1s cubic-bezier(.4,0,.6,1), svelte-z7cif2-pulse 2s cubic-bezier(.4,0,.6,1) 1s infinite;
|
118 |
+
border: 2px solid var(--color-accent);
|
119 |
+
background: transparent;
|
120 |
+
z-index: var(--layer-1);
|
121 |
+
pointer-events: none;
|
122 |
+
position: absolute;
|
123 |
+
height: 100%;
|
124 |
+
width: 100%;
|
125 |
+
border-radius: 7px;
|
126 |
+
}
|
127 |
+
#live_gen_items{
|
128 |
+
max-height: 570px;
|
129 |
+
overflow-y: scroll;
|
130 |
+
}
|
131 |
+
'''
|
132 |
+
|
133 |
+
examples = [
|
134 |
[
|
135 |
"Create a serene soundscape of a quiet beach at sunset.", # Text prompt
|
136 |
"dpmpp-2m-sde", # Sampler type
|
|
|
184 |
0.3, # Sigma min
|
185 |
500 # Sigma max
|
186 |
]
|
187 |
+
]
|
188 |
+
with gr.Blocks(theme=theme, css=css) as demo:
|
189 |
+
gr.Markdown("# Stable Audio Multiplayer Live")
|
190 |
+
gr.Markdown("Generate audio with text, share and learn from others how to best prompt this new model")
|
191 |
+
generation_history = gr.Textbox(visible=False)
|
192 |
+
list_size = gr.Number(value=PAGE_SIZE, visible=False)
|
193 |
+
with gr.Row():
|
194 |
+
with gr.Column():
|
195 |
+
prompt = gr.Textbox(label="Prompt", placeholder="Enter your text prompt here")
|
196 |
+
btn_run = gr.Button("Generate")
|
197 |
+
with gr.Accordion("Parameters", open=True):
|
198 |
+
with gr.Row():
|
199 |
+
duration = gr.Slider(0, 47, value=20, step=1, label="Duration in Seconds")
|
200 |
+
|
201 |
+
with gr.Accordion("Advanced parameters", open=False):
|
202 |
+
steps = gr.Slider(10, 150, value=80, step=10, label="Number of Diffusion Steps")
|
203 |
+
sampler_type = gr.Dropdown(["dpmpp-2m-sde", "dpmpp-3m-sde", "k-heun", "k-lms",
|
204 |
+
"k-dpmpp-2s-ancestral", "k-dpm-2", "k-dpm-fast"],
|
205 |
+
label="Sampler type", value="dpmpp-3m-sde")
|
206 |
+
with gr.Row():
|
207 |
+
cfg_scale = gr.Slider(1, 15, value=7, step=0.1, label="CFG Scale")
|
208 |
+
sigma_min = gr.Slider(0.0, 5.0, step=0.01, value=0.3, label="Sigma min")
|
209 |
+
sigma_max = gr.Slider(0.0, 1000.0, step=0.1, value=500, label="Sigma max")
|
210 |
+
with gr.Column() as output_list:
|
211 |
+
output = gr.Audio(type="filepath", label="Generated Audio")
|
212 |
+
with gr.Column(elem_id="live_gen") as community_list:
|
213 |
+
gr.Markdown("# Community generations")
|
214 |
+
with gr.Column(elem_id="live_gen_items"):
|
215 |
+
@gr.render(inputs=[generation_history, list_size])
|
216 |
+
def show_output_list(generation_history, list_size):
|
217 |
+
history_list = generation_history.split(',') if generation_history else []
|
218 |
+
history_list_latest = history_list[:list_size]
|
219 |
+
for generation in history_list_latest:
|
220 |
+
generation_prompt_file = generation.replace('.wav', '.txt')
|
221 |
+
with open(generation_prompt_file, 'r') as file:
|
222 |
+
generation_prompt = file.read()
|
223 |
+
with gr.Group():
|
224 |
+
gr.Markdown(value=f"### {generation_prompt}")
|
225 |
+
gr.Audio(value=generation)
|
226 |
+
|
227 |
+
|
228 |
+
load_more = gr.Button("Load more")
|
229 |
+
load_more.click(fn=increase_list_size, inputs=list_size, outputs=list_size)
|
230 |
+
|
231 |
+
gr.Examples(
|
232 |
+
fn=generate_audio,
|
233 |
+
examples=examples,
|
234 |
+
inputs=[prompt, sampler_type, duration, steps, cfg_scale, sigma_min, sigma_max],
|
235 |
+
outputs=output,
|
236 |
+
cache_examples="lazy"
|
237 |
+
)
|
238 |
+
gr.on(
|
239 |
+
triggers=[btn_run.click, prompt.submit],
|
240 |
+
fn=generate_audio,
|
241 |
+
inputs=[prompt, sampler_type, duration, steps, cfg_scale, sigma_min, sigma_max],
|
242 |
+
outputs=output
|
243 |
+
)
|
244 |
+
btn_run.click(
|
245 |
+
generate_audio,
|
246 |
+
inputs=[prompt, sampler_type, duration, steps, cfg_scale, sigma_min, sigma_max],
|
247 |
+
outputs=output
|
248 |
+
)
|
249 |
+
demo.load(fn=list_all_outputs, inputs=generation_history, outputs=[generation_history, community_list], every=2)
|
250 |
|
|
|
251 |
model, model_config = load_model()
|
252 |
|
253 |
+
demo.launch()
|
|
|
|