Spaces:
Runtime error
Runtime error
Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,245 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import gradio as gr
|
2 |
+
from huggingface_hub import InferenceClient
|
3 |
+
import base64
|
4 |
+
import torch
|
5 |
+
import torchaudio
|
6 |
+
from einops import rearrange
|
7 |
+
from stable_audio_tools import get_pretrained_model
|
8 |
+
from stable_audio_tools.inference.generation import generate_diffusion_cond
|
9 |
+
from diffusers import DiffusionPipeline
|
10 |
+
from huggingface_hub import InferenceClient, cached_download, hf_hub_url
|
11 |
+
from huggingface_hub import HfApi
|
12 |
+
|
13 |
+
import os
|
14 |
+
from typing import List, Dict
|
15 |
+
|
16 |
+
# Authentication
|
17 |
+
client = InferenceClient("meta-llama/Meta-Llama-3.1-8B-Instruct", token=os.environ.get("api_key"))
|
18 |
+
|
19 |
+
# Load models
|
20 |
+
device = "cuda" if torch.cuda.is_available() else "cpu"
|
21 |
+
model, model_config = get_pretrained_model("stabilityai/stable-audio-open-1.0")
|
22 |
+
sample_rate = model_config["sample_rate"]
|
23 |
+
sample_size = model_config["sample_size"]
|
24 |
+
model = model.to(device)
|
25 |
+
|
26 |
+
pipeline = DiffusionPipeline.from_pretrained("fluently/Fluently-XL-v2")
|
27 |
+
pipeline.load_lora_weights("ehristoforu/dalle-3-xl-v2")
|
28 |
+
|
29 |
+
# --- Hugging Face Spaces Storage ---
|
30 |
+
api = HfApi()
|
31 |
+
repo_id = "kvikontent/suno-ai" # Replace with your Hugging Face repository ID
|
32 |
+
|
33 |
+
# --- Global Variables ---
|
34 |
+
generated_songs = {}
|
35 |
+
|
36 |
+
# Function to generate audio (Requires GPU)
|
37 |
+
@gr.blocks
|
38 |
+
@spaces.GPU
|
39 |
+
def generate_audio(prompt: str) -> List[bytes]:
|
40 |
+
"""Generates music, image, and names a song."""
|
41 |
+
# --- Audio Generation ---
|
42 |
+
conditioning = [{
|
43 |
+
"prompt": prompt,
|
44 |
+
}]
|
45 |
+
|
46 |
+
output = generate_diffusion_cond(
|
47 |
+
model,
|
48 |
+
conditioning=conditioning,
|
49 |
+
sample_size=sample_size,
|
50 |
+
device=device
|
51 |
+
)
|
52 |
+
|
53 |
+
output = rearrange(output, "b d n -> d (b n)")
|
54 |
+
|
55 |
+
# Peak normalize, clip, convert to int16, and save to file
|
56 |
+
output = output.to(torch.float32).div(torch.max(torch.abs(output))).clamp(-1, 1).mul(32767).to(torch.int16).cpu()
|
57 |
+
|
58 |
+
# Save audio to memory
|
59 |
+
buffer = BytesIO()
|
60 |
+
torchaudio.save(buffer, output, sample_rate)
|
61 |
+
audio_data = buffer.getvalue()
|
62 |
+
|
63 |
+
# --- Image Generation ---
|
64 |
+
image = pipeline(prompt).images[0]
|
65 |
+
buffer = BytesIO()
|
66 |
+
image.save(buffer, format='png')
|
67 |
+
image_data = buffer.getvalue()
|
68 |
+
|
69 |
+
# --- Name Generation ---
|
70 |
+
for message in client.chat_completion(
|
71 |
+
messages=[{"role": "user", "content": "Name the song based on this prompt: " + prompt}],
|
72 |
+
max_tokens=500,
|
73 |
+
stream=True,
|
74 |
+
):
|
75 |
+
song_name = message.choices[0].delta.content
|
76 |
+
|
77 |
+
return audio_data, image_data, song_name
|
78 |
+
|
79 |
+
# Function to download generated audio and image
|
80 |
+
def download_audio_image(audio_data, image_data, song_name):
|
81 |
+
"""Downloads generated audio and image."""
|
82 |
+
audio_bytes = base64.b64encode(audio_data).decode('utf-8')
|
83 |
+
image_bytes = base64.b64encode(image_data).decode('utf-8')
|
84 |
+
|
85 |
+
audio_url = f"data:audio/wav;base64,{audio_bytes}"
|
86 |
+
image_url = f"data:image/png;base64,{image_bytes}"
|
87 |
+
|
88 |
+
return audio_url, image_url, song_name
|
89 |
+
|
90 |
+
# Function to make a song public
|
91 |
+
def make_public(song_id, audio_data, image_data, song_name, user_id):
|
92 |
+
"""Makes a song public."""
|
93 |
+
generated_songs[song_id]["public"] = True
|
94 |
+
|
95 |
+
# Save the song data to Hugging Face Spaces
|
96 |
+
api.upload_file(
|
97 |
+
path="audio.wav",
|
98 |
+
path_in_repo=f"songs/{song_id}/audio.wav",
|
99 |
+
repo_id=repo_id,
|
100 |
+
repo_type="space",
|
101 |
+
data=audio_data
|
102 |
+
)
|
103 |
+
api.upload_file(
|
104 |
+
path="image.png",
|
105 |
+
path_in_repo=f"songs/{song_id}/image.png",
|
106 |
+
repo_id=repo_id,
|
107 |
+
repo_type="space",
|
108 |
+
data=image_data
|
109 |
+
)
|
110 |
+
# Save the song name as a text file
|
111 |
+
with open(f"song_name.txt", "w") as f:
|
112 |
+
f.write(song_name)
|
113 |
+
api.upload_file(
|
114 |
+
path="song_name.txt",
|
115 |
+
path_in_repo=f"songs/{song_id}/song_name.txt",
|
116 |
+
repo_id=repo_id,
|
117 |
+
repo_type="space",
|
118 |
+
)
|
119 |
+
|
120 |
+
return generated_songs
|
121 |
+
|
122 |
+
# Function to fetch songs from Hugging Face Spaces
|
123 |
+
def fetch_songs(user_id=None):
|
124 |
+
"""Fetches songs from Hugging Face Spaces."""
|
125 |
+
songs = {}
|
126 |
+
files = api.list_repo_files(repo_id=repo_id, repo_type="space")
|
127 |
+
for file in files:
|
128 |
+
if file["path"].startswith("songs"):
|
129 |
+
song_id = file["path"].split("/")[1]
|
130 |
+
if song_id not in songs:
|
131 |
+
songs[song_id] = {}
|
132 |
+
|
133 |
+
if "audio.wav" in file["path"]:
|
134 |
+
# Fetch audio data
|
135 |
+
audio_data = api.download_file(repo_id=repo_id, repo_type="space", revision="main", path=file["path"])
|
136 |
+
songs[song_id]["audio"] = audio_data
|
137 |
+
|
138 |
+
if "image.png" in file["path"]:
|
139 |
+
# Fetch image data
|
140 |
+
image_data = api.download_file(repo_id=repo_id, repo_type="space", revision="main", path=file["path"])
|
141 |
+
songs[song_id]["image"] = image_data
|
142 |
+
|
143 |
+
if "song_name.txt" in file["path"]:
|
144 |
+
# Fetch song name data
|
145 |
+
with open("song_name.txt", "wb") as f:
|
146 |
+
f.write(api.download_file(repo_id=repo_id, repo_type="space", revision="main", path=file["path"]))
|
147 |
+
with open("song_name.txt", "r") as f:
|
148 |
+
song_name = f.read()
|
149 |
+
songs[song_id]["name"] = song_name
|
150 |
+
|
151 |
+
# Extract the public/private status and user ID from the file name (if available)
|
152 |
+
# ... (Implement logic here based on how you store this information)
|
153 |
+
# ...
|
154 |
+
|
155 |
+
return songs
|
156 |
+
|
157 |
+
# --- User Interface ---
|
158 |
+
with gr.Blocks() as demo:
|
159 |
+
gr.Markdown("## Neon Synth Music Generator")
|
160 |
+
|
161 |
+
# Input area
|
162 |
+
prompt_input = gr.Textbox(label="Prompt", placeholder="e.g., 128 BPM tech house drum loop")
|
163 |
+
generate_button = gr.Button("Generate")
|
164 |
+
|
165 |
+
# Output area
|
166 |
+
generated_audio = gr.Audio(label="Generated Audio", playable=True, source="upload")
|
167 |
+
generated_image = gr.Image(label="Generated Image")
|
168 |
+
song_name = gr.Textbox(label="Song Name")
|
169 |
+
make_public_button = gr.Button("Make Public")
|
170 |
+
|
171 |
+
# User authentication
|
172 |
+
login_button = gr.Button("Login")
|
173 |
+
logout_button = gr.Button("Logout", visible=False)
|
174 |
+
user_name = gr.Textbox(label="Username", interactive=False, visible=False)
|
175 |
+
|
176 |
+
# Feed area
|
177 |
+
public_feed = gr.Gallery(label="Public Feed", show_label=False, elem_id="public-feed")
|
178 |
+
user_feed = gr.Gallery(label="Your Feed", show_label=False, elem_id="user-feed")
|
179 |
+
|
180 |
+
# --- Event Handlers ---
|
181 |
+
generate_button.click(fn=generate_audio, inputs=prompt_input, outputs=[generated_audio, generated_image, song_name])
|
182 |
+
make_public_button.click(fn=make_public, inputs=[gr.State(generated_songs), generated_audio, generated_image, song_name, gr.State(user_name)], outputs=[gr.State(generated_songs)], show_error=False)
|
183 |
+
login_button.click(fn=lambda: "YourUsername", inputs=[], outputs=[user_name], show_error=False)
|
184 |
+
logout_button.click(fn=lambda: "", inputs=[], outputs=[user_name], show_error=False)
|
185 |
+
login_button.click(fn=lambda: gr.update(visible=False), inputs=[], outputs=login_button, show_error=False)
|
186 |
+
login_button.click(fn=lambda: gr.update(visible=True), inputs=[], outputs=logout_button, show_error=False)
|
187 |
+
login_button.click(fn=lambda: gr.update(visible=True), inputs=[], outputs=user_name, show_error=False)
|
188 |
+
logout_button.click(fn=lambda: gr.update(visible=True), inputs=[], outputs=login_button, show_error=False)
|
189 |
+
logout_button.click(fn=lambda: gr.update(visible=False), inputs=[], outputs=logout_button, show_error=False)
|
190 |
+
logout_button.click(fn=lambda: gr.update(visible=False), inputs=[], outputs=user_name, show_error=False)
|
191 |
+
|
192 |
+
# --- Update the feed ---
|
193 |
+
generated_audio.change(fn=download_audio_image, inputs=[generated_audio, generated_image, song_name], outputs=[generated_audio, generated_image, song_name], show_error=False)
|
194 |
+
generated_audio.change(
|
195 |
+
fn=lambda audio_data, image_data, song_name, user_name: [
|
196 |
+
{"audio": audio_data, "image": image_data, "name": song_name, "public": False, "user": user_name}
|
197 |
+
],
|
198 |
+
inputs=[generated_audio, generated_image, song_name, user_name],
|
199 |
+
outputs=[gr.State(generated_songs)],
|
200 |
+
show_error=False,
|
201 |
+
)
|
202 |
+
|
203 |
+
# Refresh the feed when a new song is added
|
204 |
+
generated_songs.change(
|
205 |
+
fn=lambda generated_songs: [
|
206 |
+
[gr.update(value=download_audio_image(s["audio"], s["image"], s["name"])) for s in generated_songs.values() if s["public"]],
|
207 |
+
[gr.update(value=download_audio_image(s["audio"], s["image"], s["name"])) for s in generated_songs.values() if not s["public"] and s["user"] == user_name]
|
208 |
+
],
|
209 |
+
inputs=[gr.State(generated_songs)],
|
210 |
+
outputs=[public_feed, user_feed],
|
211 |
+
show_error=False,
|
212 |
+
)
|
213 |
+
|
214 |
+
# Fetch and display the feeds
|
215 |
+
demo.load(
|
216 |
+
fn=lambda: [
|
217 |
+
[gr.update(value=download_audio_image(s["audio"], s["image"], s["name"])) for s in fetch_songs().values() if s["public"]],
|
218 |
+
[gr.update(value=download_audio_image(s["audio"], s["image"], s["name"])) for s in fetch_songs(user_name).values() if not s["public"]]
|
219 |
+
],
|
220 |
+
outputs=[public_feed, user_feed],
|
221 |
+
show_error=False,
|
222 |
+
)
|
223 |
+
|
224 |
+
# --- Layout ---
|
225 |
+
with gr.Row():
|
226 |
+
with gr.Column():
|
227 |
+
prompt_input
|
228 |
+
generate_button
|
229 |
+
login_button
|
230 |
+
logout_button
|
231 |
+
user_name
|
232 |
+
with gr.Column():
|
233 |
+
generated_audio
|
234 |
+
generated_image
|
235 |
+
song_name
|
236 |
+
make_public_button
|
237 |
+
|
238 |
+
with gr.Row():
|
239 |
+
with gr.Column():
|
240 |
+
public_feed
|
241 |
+
with gr.Column():
|
242 |
+
user_feed
|
243 |
+
|
244 |
+
# Run the Gradio interface
|
245 |
+
demo.launch()
|