Spaces:
Runtime error
Runtime error
Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,116 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# app.py
|
2 |
+
import gradio as gr
|
3 |
+
import torch
|
4 |
+
import transformers
|
5 |
+
import librosa
|
6 |
+
import numpy as np
|
7 |
+
import os
|
8 |
+
|
9 |
+
class UltravoxInterface:
|
10 |
+
def __init__(self):
|
11 |
+
"""Initialize the Ultravox model and settings"""
|
12 |
+
print("Loading Ultravox model... This may take a few minutes...")
|
13 |
+
self.pipe = transformers.pipeline(
|
14 |
+
model='fixie-ai/ultravox-v0_4',
|
15 |
+
trust_remote_code=True,
|
16 |
+
device=0 if torch.cuda.is_available() else -1
|
17 |
+
)
|
18 |
+
print("Model loaded successfully!")
|
19 |
+
|
20 |
+
# Default system prompt
|
21 |
+
self.default_prompt = "You are a friendly and helpful character. You love to answer questions for people."
|
22 |
+
|
23 |
+
def process_audio(self, audio_path, custom_prompt=None):
|
24 |
+
"""Process audio input and return model response"""
|
25 |
+
try:
|
26 |
+
# Load and preprocess audio
|
27 |
+
audio, sr = librosa.load(audio_path, sr=16000)
|
28 |
+
|
29 |
+
# Prepare conversation turns
|
30 |
+
turns = [
|
31 |
+
{
|
32 |
+
"role": "system",
|
33 |
+
"content": custom_prompt if custom_prompt else self.default_prompt
|
34 |
+
}
|
35 |
+
]
|
36 |
+
|
37 |
+
# Get model response
|
38 |
+
result = self.pipe(
|
39 |
+
{
|
40 |
+
'audio': audio,
|
41 |
+
'turns': turns,
|
42 |
+
'sampling_rate': sr
|
43 |
+
},
|
44 |
+
max_new_tokens=30
|
45 |
+
)
|
46 |
+
|
47 |
+
# Handle different response formats
|
48 |
+
if isinstance(result, str):
|
49 |
+
return result
|
50 |
+
elif isinstance(result, list):
|
51 |
+
return result[0] if result else "No response generated"
|
52 |
+
elif isinstance(result, dict):
|
53 |
+
return result.get('generated_text', "No response generated")
|
54 |
+
else:
|
55 |
+
return str(result)
|
56 |
+
|
57 |
+
except Exception as e:
|
58 |
+
return f"Error processing audio: {str(e)}"
|
59 |
+
|
60 |
+
def create_interface(self):
|
61 |
+
"""Create and configure the Gradio interface"""
|
62 |
+
|
63 |
+
with gr.Blocks(title="Ultravox Voice Assistant", theme=gr.themes.Soft(
|
64 |
+
primary_hue="orange",
|
65 |
+
secondary_hue="gray",
|
66 |
+
)) as interface:
|
67 |
+
gr.Markdown("# 🎙️ Ultravox Voice Assistant")
|
68 |
+
gr.Markdown("Speak into the microphone and get AI-generated responses!")
|
69 |
+
|
70 |
+
with gr.Row():
|
71 |
+
with gr.Column():
|
72 |
+
audio_input = gr.Audio(
|
73 |
+
label="Speak here",
|
74 |
+
sources=["microphone"],
|
75 |
+
type="filepath"
|
76 |
+
)
|
77 |
+
|
78 |
+
system_prompt = gr.Textbox(
|
79 |
+
label="System Prompt (Optional)",
|
80 |
+
placeholder="Enter custom system prompt or leave empty for default",
|
81 |
+
value=self.default_prompt
|
82 |
+
)
|
83 |
+
|
84 |
+
submit_btn = gr.Button(
|
85 |
+
"Process Audio",
|
86 |
+
variant="primary"
|
87 |
+
)
|
88 |
+
|
89 |
+
with gr.Column():
|
90 |
+
output_text = gr.Textbox(
|
91 |
+
label="AI Response",
|
92 |
+
lines=5,
|
93 |
+
placeholder="AI response will appear here..."
|
94 |
+
)
|
95 |
+
|
96 |
+
submit_btn.click(
|
97 |
+
fn=self.process_audio,
|
98 |
+
inputs=[audio_input, system_prompt],
|
99 |
+
outputs=output_text
|
100 |
+
)
|
101 |
+
|
102 |
+
gr.Markdown("""
|
103 |
+
## How to use:
|
104 |
+
1. Click the microphone icon and allow browser access
|
105 |
+
2. Speak your question or prompt
|
106 |
+
3. Click 'Stop' when finished
|
107 |
+
4. Click 'Process Audio' to get the AI response
|
108 |
+
|
109 |
+
## Note:
|
110 |
+
First-time loading may take a few minutes as the model is downloaded.
|
111 |
+
""")
|
112 |
+
|
113 |
+
return interface
|
114 |
+
|
115 |
+
app = UltravoxInterface()
|
116 |
+
interface = app.create_interface()
|