Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
|
@@ -8,67 +8,102 @@ import gradio as gr
|
|
| 8 |
import os
|
| 9 |
from huggingface_hub import login
|
| 10 |
|
|
|
|
|
|
|
|
|
|
| 11 |
# Authentifizierung
|
| 12 |
if os.environ.get("HUGGING_FACE_HUB_TOKEN"):
|
| 13 |
-
|
| 14 |
-
|
|
|
|
|
|
|
|
|
|
| 15 |
|
| 16 |
-
@spaces.GPU(duration=
|
| 17 |
def generate_audio(prompt, duration=10, steps=50, cfg_scale=7):
|
| 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 |
-
gr.
|
| 66 |
-
|
| 67 |
-
|
| 68 |
-
|
| 69 |
-
|
| 70 |
-
|
| 71 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 72 |
|
| 73 |
if __name__ == "__main__":
|
| 74 |
-
|
|
|
|
| 8 |
import os
|
| 9 |
from huggingface_hub import login
|
| 10 |
|
| 11 |
+
# Ordner für temporäre Dateien erstellen
|
| 12 |
+
os.makedirs('static', exist_ok=True)
|
| 13 |
+
|
| 14 |
# Authentifizierung
|
| 15 |
if os.environ.get("HUGGING_FACE_HUB_TOKEN"):
|
| 16 |
+
token = os.environ["HUGGING_FACE_HUB_TOKEN"].strip()
|
| 17 |
+
try:
|
| 18 |
+
login(token=token, add_to_git_credential=True)
|
| 19 |
+
except Exception as e:
|
| 20 |
+
print(f"Warnung: Login fehlgeschlagen - {str(e)}")
|
| 21 |
|
| 22 |
+
@spaces.GPU(duration=300)
|
| 23 |
def generate_audio(prompt, duration=10, steps=50, cfg_scale=7):
|
| 24 |
+
try:
|
| 25 |
+
device = "cuda" if torch.cuda.is_available() else "cpu"
|
| 26 |
+
|
| 27 |
+
# Modell laden und zum Gerät verschieben
|
| 28 |
+
model, model_config = get_pretrained_model("stabilityai/stable-audio-open-1.0")
|
| 29 |
+
model = model.to(device)
|
| 30 |
+
|
| 31 |
+
sample_rate = model_config["sample_rate"]
|
| 32 |
+
sample_size = model_config["sample_size"]
|
| 33 |
|
| 34 |
+
# Konditionierung einrichten
|
| 35 |
+
conditioning = [{
|
| 36 |
+
"prompt": prompt,
|
| 37 |
+
"seconds_start": 0,
|
| 38 |
+
"seconds_total": duration # Keine Begrenzung mehr
|
| 39 |
+
}]
|
| 40 |
|
| 41 |
+
# Audio generieren mit anpassbaren Parametern
|
| 42 |
+
output = generate_diffusion_cond(
|
| 43 |
+
model,
|
| 44 |
+
steps=steps, # Keine Begrenzung mehr
|
| 45 |
+
cfg_scale=cfg_scale,
|
| 46 |
+
conditioning=conditioning,
|
| 47 |
+
sample_size=sample_size,
|
| 48 |
+
sigma_min=0.3,
|
| 49 |
+
sigma_max=500,
|
| 50 |
+
sampler_type="dpmpp-3m-sde",
|
| 51 |
+
device=device
|
| 52 |
+
)
|
| 53 |
|
| 54 |
+
# Audio verarbeiten
|
| 55 |
+
output = rearrange(output, "b d n -> d (b n)")
|
| 56 |
+
output = output.to(torch.float32).div(torch.max(torch.abs(output))).clamp(-1, 1).mul(32767).to(torch.int16).cpu()
|
| 57 |
|
| 58 |
+
# Audio speichern
|
| 59 |
+
output_path = "static/generated_audio.wav"
|
| 60 |
+
torchaudio.save(output_path, output, model_config["sample_rate"])
|
| 61 |
+
|
| 62 |
+
return output_path
|
| 63 |
|
| 64 |
+
except Exception as e:
|
| 65 |
+
print(f"Fehler bei der Audiogenerierung: {str(e)}")
|
| 66 |
+
raise e
|
| 67 |
|
| 68 |
+
# Benutzerdefiniertes CSS für besseres Aussehen
|
| 69 |
+
custom_css = """
|
| 70 |
+
body { background-color: #f6f6f6; }
|
| 71 |
+
.gradio-container { max-width: 800px; margin: auto; }
|
| 72 |
+
"""
|
| 73 |
|
| 74 |
+
# Gradio Interface mit Blocks
|
| 75 |
+
with gr.Blocks(css=custom_css) as demo:
|
| 76 |
+
gr.Markdown("# Stable Audio Generator")
|
| 77 |
+
gr.Markdown("Generieren Sie Audio aus Textbeschreibungen mit Stable Audio 1.0")
|
| 78 |
+
|
| 79 |
+
with gr.Row():
|
| 80 |
+
with gr.Column():
|
| 81 |
+
prompt = gr.Textbox(
|
| 82 |
+
label="Prompt",
|
| 83 |
+
placeholder="Beschreiben Sie den gewünschten Sound..."
|
| 84 |
+
)
|
| 85 |
+
duration = gr.Slider(
|
| 86 |
+
minimum=1, maximum=30, value=10,
|
| 87 |
+
step=1, label="Dauer (Sekunden)"
|
| 88 |
+
)
|
| 89 |
+
steps = gr.Slider(
|
| 90 |
+
minimum=20, maximum=100, value=50,
|
| 91 |
+
step=5, label="Anzahl der Schritte"
|
| 92 |
+
)
|
| 93 |
+
cfg_scale = gr.Slider(
|
| 94 |
+
minimum=1, maximum=15, value=7,
|
| 95 |
+
step=0.5, label="CFG Scale"
|
| 96 |
+
)
|
| 97 |
+
generate_btn = gr.Button("Generieren")
|
| 98 |
+
|
| 99 |
+
with gr.Column():
|
| 100 |
+
output = gr.Audio(label="Generiertes Audio", type="filepath")
|
| 101 |
+
|
| 102 |
+
generate_btn.click(
|
| 103 |
+
fn=generate_audio,
|
| 104 |
+
inputs=[prompt, duration, steps, cfg_scale],
|
| 105 |
+
outputs=output
|
| 106 |
+
)
|
| 107 |
|
| 108 |
if __name__ == "__main__":
|
| 109 |
+
demo.launch()
|