Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
@@ -4,9 +4,8 @@ import torchaudio
|
|
4 |
import numpy as np
|
5 |
from transformers import AutoProcessor, SeamlessM4Tv2Model
|
6 |
from datetime import datetime
|
7 |
-
import time
|
8 |
|
9 |
-
class
|
10 |
def __init__(self, model_name: str = "facebook/seamless-m4t-v2-large"):
|
11 |
self.processor = AutoProcessor.from_pretrained(model_name)
|
12 |
self.model = SeamlessM4Tv2Model.from_pretrained(model_name)
|
@@ -25,131 +24,201 @@ class JarvisTranslator:
|
|
25 |
}
|
26 |
|
27 |
def translate(self, text: str, src_lang: str, tgt_lang: str) -> tuple[int, np.ndarray]:
|
28 |
-
|
29 |
-
|
30 |
-
|
|
|
|
|
|
|
31 |
|
32 |
-
|
33 |
-
|
34 |
-
|
35 |
-
|
36 |
-
|
37 |
-
|
38 |
-
|
39 |
-
|
40 |
-
border-radius: 10px;
|
41 |
-
border: 2px solid #00ffff;
|
42 |
-
}
|
43 |
-
|
44 |
-
#status-circle {
|
45 |
-
width: 150px;
|
46 |
-
height: 150px;
|
47 |
-
border: 4px solid #00ffff;
|
48 |
-
border-radius: 50%;
|
49 |
-
margin: 20px auto;
|
50 |
-
position: relative;
|
51 |
-
animation: pulse 2s infinite;
|
52 |
-
}
|
53 |
-
|
54 |
-
@keyframes pulse {
|
55 |
-
0% { box-shadow: 0 0 0 0 rgba(0, 255, 255, 0.4); }
|
56 |
-
70% { box-shadow: 0 0 0 20px rgba(0, 255, 255, 0); }
|
57 |
-
100% { box-shadow: 0 0 0 0 rgba(0, 255, 255, 0); }
|
58 |
-
}
|
59 |
-
|
60 |
-
.custom-button {
|
61 |
-
background-color: transparent !important;
|
62 |
-
border: 2px solid #00ffff !important;
|
63 |
-
color: #00ffff !important;
|
64 |
-
font-family: 'Courier New', monospace !important;
|
65 |
-
}
|
66 |
-
|
67 |
-
.custom-button:hover {
|
68 |
-
background-color: rgba(0, 255, 255, 0.1) !important;
|
69 |
-
}
|
70 |
-
|
71 |
-
.status-text {
|
72 |
-
color: #00ffff;
|
73 |
-
text-align: center;
|
74 |
-
font-size: 1.2em;
|
75 |
-
margin: 10px 0;
|
76 |
-
}
|
77 |
-
|
78 |
-
.time-display {
|
79 |
-
position: absolute;
|
80 |
-
top: 10px;
|
81 |
-
right: 10px;
|
82 |
-
color: #00ffff;
|
83 |
-
font-family: 'Courier New', monospace;
|
84 |
-
}
|
85 |
-
"""
|
86 |
|
87 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
88 |
|
89 |
-
|
90 |
-
|
|
|
|
|
|
|
91 |
|
92 |
-
|
93 |
-
|
94 |
-
|
95 |
-
|
96 |
-
|
97 |
-
|
98 |
-
|
99 |
-
|
100 |
-
|
101 |
-
return None, f"Error: {str(e)}"
|
102 |
|
103 |
-
|
104 |
-
|
105 |
-
|
106 |
-
|
107 |
-
|
108 |
-
|
109 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
110 |
|
|
|
111 |
with gr.Row():
|
112 |
text_input = gr.Textbox(
|
113 |
label="Command Input",
|
114 |
placeholder="Enter text to translate...",
|
|
|
115 |
lines=3
|
116 |
)
|
117 |
-
|
118 |
with gr.Row():
|
119 |
src_lang = gr.Dropdown(
|
120 |
choices=list(translator.language_codes.keys()),
|
121 |
value="English",
|
122 |
-
label="Source Language"
|
|
|
123 |
)
|
124 |
tgt_lang = gr.Dropdown(
|
125 |
choices=list(translator.language_codes.keys()),
|
126 |
value="Spanish",
|
127 |
-
label="Target Language"
|
|
|
128 |
)
|
129 |
-
|
130 |
-
|
131 |
-
|
132 |
-
|
|
|
|
|
|
|
133 |
audio_output = gr.Audio(
|
134 |
-
label="Translated
|
135 |
type="numpy"
|
136 |
)
|
137 |
|
138 |
-
#
|
139 |
-
|
140 |
-
|
141 |
-
|
142 |
-
|
143 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
144 |
|
145 |
-
|
146 |
-
|
147 |
-
|
148 |
-
|
|
|
149 |
)
|
150 |
-
|
151 |
return demo
|
152 |
|
153 |
if __name__ == "__main__":
|
154 |
-
demo =
|
|
|
155 |
demo.launch()
|
|
|
4 |
import numpy as np
|
5 |
from transformers import AutoProcessor, SeamlessM4Tv2Model
|
6 |
from datetime import datetime
|
|
|
7 |
|
8 |
+
class SeamlessTranslator:
|
9 |
def __init__(self, model_name: str = "facebook/seamless-m4t-v2-large"):
|
10 |
self.processor = AutoProcessor.from_pretrained(model_name)
|
11 |
self.model = SeamlessM4Tv2Model.from_pretrained(model_name)
|
|
|
24 |
}
|
25 |
|
26 |
def translate(self, text: str, src_lang: str, tgt_lang: str) -> tuple[int, np.ndarray]:
|
27 |
+
try:
|
28 |
+
inputs = self.processor(text=text, src_lang=self.language_codes[src_lang], return_tensors="pt")
|
29 |
+
audio_array = self.model.generate(**inputs, tgt_lang=self.language_codes[tgt_lang])[0].cpu().numpy().squeeze()
|
30 |
+
return self.sample_rate, audio_array
|
31 |
+
except Exception as e:
|
32 |
+
raise gr.Error(f"Translation failed: {str(e)}")
|
33 |
|
34 |
+
# Custom CSS for Jarvis theme
|
35 |
+
css = """
|
36 |
+
#jarvis-interface {
|
37 |
+
background-color: black !important;
|
38 |
+
background-image: radial-gradient(circle at center, #00303030 0%, #00000080 100%);
|
39 |
+
min-height: 100vh;
|
40 |
+
font-family: 'Courier New', monospace;
|
41 |
+
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
42 |
|
43 |
+
#status-ring {
|
44 |
+
width: 300px;
|
45 |
+
height: 300px;
|
46 |
+
border: 4px solid #00ffff;
|
47 |
+
border-radius: 50%;
|
48 |
+
margin: 20px auto;
|
49 |
+
position: relative;
|
50 |
+
animation: pulse 2s infinite;
|
51 |
+
display: flex;
|
52 |
+
align-items: center;
|
53 |
+
justify-content: center;
|
54 |
+
}
|
55 |
|
56 |
+
@keyframes pulse {
|
57 |
+
0% { box-shadow: 0 0 0 0 rgba(0, 255, 255, 0.4); }
|
58 |
+
70% { box-shadow: 0 0 0 20px rgba(0, 255, 255, 0); }
|
59 |
+
100% { box-shadow: 0 0 0 0 rgba(0, 255, 255, 0); }
|
60 |
+
}
|
61 |
|
62 |
+
#inner-ring {
|
63 |
+
width: 200px;
|
64 |
+
height: 200px;
|
65 |
+
border: 2px solid #00ffff;
|
66 |
+
border-radius: 50%;
|
67 |
+
display: flex;
|
68 |
+
align-items: center;
|
69 |
+
justify-content: center;
|
70 |
+
}
|
|
|
71 |
|
72 |
+
#core {
|
73 |
+
width: 100px;
|
74 |
+
height: 100px;
|
75 |
+
border: 3px solid #00ffff;
|
76 |
+
border-radius: 50%;
|
77 |
+
background-color: black;
|
78 |
+
display: flex;
|
79 |
+
align-items: center;
|
80 |
+
justify-content: center;
|
81 |
+
color: #00ffff;
|
82 |
+
text-align: center;
|
83 |
+
padding: 10px;
|
84 |
+
}
|
85 |
+
|
86 |
+
.jarvis-textbox {
|
87 |
+
background-color: black !important;
|
88 |
+
border: 2px solid #00ffff !important;
|
89 |
+
color: #00ffff !important;
|
90 |
+
font-family: 'Courier New', monospace !important;
|
91 |
+
}
|
92 |
+
|
93 |
+
.jarvis-button {
|
94 |
+
background-color: transparent !important;
|
95 |
+
border: 2px solid #00ffff !important;
|
96 |
+
color: #00ffff !important;
|
97 |
+
font-family: 'Courier New', monospace !important;
|
98 |
+
}
|
99 |
+
|
100 |
+
.jarvis-button:hover {
|
101 |
+
background-color: rgba(0, 255, 255, 0.1) !important;
|
102 |
+
}
|
103 |
+
|
104 |
+
.status-box {
|
105 |
+
background-color: black !important;
|
106 |
+
border: 2px solid #00ffff !important;
|
107 |
+
color: #00ffff !important;
|
108 |
+
padding: 10px !important;
|
109 |
+
border-radius: 5px !important;
|
110 |
+
margin: 5px !important;
|
111 |
+
text-align: center !important;
|
112 |
+
}
|
113 |
+
"""
|
114 |
+
|
115 |
+
def create_interface():
|
116 |
+
translator = SeamlessTranslator()
|
117 |
+
|
118 |
+
def translate_text(text: str, src_lang: str, tgt_lang: str, progress=gr.Progress()):
|
119 |
+
progress(0, desc="Initializing...")
|
120 |
+
progress(0.3, desc="Processing text...")
|
121 |
+
sample_rate, audio = translator.translate(text, src_lang, tgt_lang)
|
122 |
+
progress(0.7, desc="Generating audio...")
|
123 |
+
progress(1.0, desc="Complete!")
|
124 |
+
return audio
|
125 |
+
|
126 |
+
with gr.Blocks(css=css, title="J.A.R.V.I.S Translator") as demo:
|
127 |
+
gr.Markdown(
|
128 |
+
"""
|
129 |
+
# J.A.R.V.I.S TRANSLATION SYSTEM
|
130 |
+
### Powered by SeamlessM4T
|
131 |
+
"""
|
132 |
+
)
|
133 |
+
|
134 |
+
# Jarvis interface container
|
135 |
+
with gr.Column(elem_id="jarvis-interface"):
|
136 |
+
# Status Ring
|
137 |
+
gr.HTML("""
|
138 |
+
<div id="status-ring">
|
139 |
+
<div id="inner-ring">
|
140 |
+
<div id="core">
|
141 |
+
<div>JARVIS</div>
|
142 |
+
<div>ACTIVE</div>
|
143 |
+
</div>
|
144 |
+
</div>
|
145 |
+
</div>
|
146 |
+
""")
|
147 |
|
148 |
+
# Input controls
|
149 |
with gr.Row():
|
150 |
text_input = gr.Textbox(
|
151 |
label="Command Input",
|
152 |
placeholder="Enter text to translate...",
|
153 |
+
elem_classes=["jarvis-textbox"],
|
154 |
lines=3
|
155 |
)
|
156 |
+
|
157 |
with gr.Row():
|
158 |
src_lang = gr.Dropdown(
|
159 |
choices=list(translator.language_codes.keys()),
|
160 |
value="English",
|
161 |
+
label="Source Language",
|
162 |
+
elem_classes=["jarvis-textbox"]
|
163 |
)
|
164 |
tgt_lang = gr.Dropdown(
|
165 |
choices=list(translator.language_codes.keys()),
|
166 |
value="Spanish",
|
167 |
+
label="Target Language",
|
168 |
+
elem_classes=["jarvis-textbox"]
|
169 |
)
|
170 |
+
|
171 |
+
translate_btn = gr.Button(
|
172 |
+
"▶ EXECUTE TRANSLATION",
|
173 |
+
elem_classes=["jarvis-button"]
|
174 |
+
)
|
175 |
+
|
176 |
+
# Output audio
|
177 |
audio_output = gr.Audio(
|
178 |
+
label="Translated Speech",
|
179 |
type="numpy"
|
180 |
)
|
181 |
|
182 |
+
# Status boxes
|
183 |
+
with gr.Row():
|
184 |
+
with gr.Column():
|
185 |
+
gr.Markdown(
|
186 |
+
"""
|
187 |
+
<div class="status-box">
|
188 |
+
SYSTEM STATUS<br>
|
189 |
+
<strong>ACTIVE</strong>
|
190 |
+
</div>
|
191 |
+
"""
|
192 |
+
)
|
193 |
+
with gr.Column():
|
194 |
+
gr.Markdown(
|
195 |
+
"""
|
196 |
+
<div class="status-box">
|
197 |
+
AUDIO SYSTEM<br>
|
198 |
+
<strong>READY</strong>
|
199 |
+
</div>
|
200 |
+
"""
|
201 |
+
)
|
202 |
+
with gr.Column():
|
203 |
+
gr.Markdown(
|
204 |
+
"""
|
205 |
+
<div class="status-box">
|
206 |
+
TRANSLATION<br>
|
207 |
+
<strong>ONLINE</strong>
|
208 |
+
</div>
|
209 |
+
"""
|
210 |
+
)
|
211 |
|
212 |
+
# Event handler
|
213 |
+
translate_btn.click(
|
214 |
+
fn=translate_text,
|
215 |
+
inputs=[text_input, src_lang, tgt_lang],
|
216 |
+
outputs=audio_output
|
217 |
)
|
218 |
+
|
219 |
return demo
|
220 |
|
221 |
if __name__ == "__main__":
|
222 |
+
demo = create_interface()
|
223 |
+
demo.queue()
|
224 |
demo.launch()
|