Spaces:
Sleeping
Sleeping
Commit Β·
6497d65
1
Parent(s): ec35154
Update design
Browse files
app.py
CHANGED
|
@@ -13,143 +13,154 @@ def clean_transcript(text: str) -> str:
|
|
| 13 |
|
| 14 |
def transcribe(audio, model_choice):
|
| 15 |
if audio is None:
|
| 16 |
-
return "
|
| 17 |
|
| 18 |
try:
|
| 19 |
-
if model_choice == "
|
| 20 |
result = whisper_asr(audio)
|
| 21 |
-
return f"
|
| 22 |
else:
|
| 23 |
result = wav2vec_asr(audio)
|
| 24 |
cleaned_text = clean_transcript(result["text"])
|
| 25 |
-
return f"
|
| 26 |
except Exception as e:
|
| 27 |
-
return f"
|
| 28 |
|
| 29 |
def clear_all():
|
| 30 |
-
return None, "
|
| 31 |
|
| 32 |
-
# Custom CSS for
|
| 33 |
custom_css = """
|
| 34 |
/* Global Styles */
|
| 35 |
.gradio-container {
|
| 36 |
-
background:
|
| 37 |
-
font-family: '
|
|
|
|
| 38 |
}
|
| 39 |
|
| 40 |
/* Header Styling */
|
| 41 |
.header-section {
|
| 42 |
-
background:
|
| 43 |
-
border
|
|
|
|
| 44 |
padding: 30px;
|
| 45 |
margin-bottom: 25px;
|
| 46 |
-
box-shadow: 0
|
| 47 |
-
backdrop-filter: blur(10px);
|
| 48 |
-
border: 1px solid rgba(255, 255, 255, 0.2);
|
| 49 |
}
|
| 50 |
|
| 51 |
.header-section h1 {
|
| 52 |
-
|
| 53 |
-
-
|
| 54 |
-
-
|
| 55 |
-
font-size: 2.5em !important;
|
| 56 |
-
font-weight: 700 !important;
|
| 57 |
text-align: center;
|
| 58 |
margin-bottom: 15px;
|
|
|
|
|
|
|
| 59 |
}
|
| 60 |
|
| 61 |
.header-section p {
|
| 62 |
-
color: #
|
| 63 |
font-size: 1.1em;
|
| 64 |
text-align: center;
|
| 65 |
-
margin: 0;
|
| 66 |
-
line-height: 1.
|
| 67 |
}
|
| 68 |
|
| 69 |
/* Main Content Cards */
|
| 70 |
.input-card, .output-card {
|
| 71 |
-
background:
|
| 72 |
-
border
|
|
|
|
| 73 |
padding: 25px;
|
| 74 |
-
box-shadow: 0
|
| 75 |
-
backdrop-filter: blur(10px);
|
| 76 |
-
border: 1px solid rgba(255, 255, 255, 0.2);
|
| 77 |
margin-bottom: 20px;
|
| 78 |
}
|
| 79 |
|
| 80 |
/* Button Styling */
|
| 81 |
.primary-button {
|
| 82 |
-
background:
|
| 83 |
-
border:
|
| 84 |
-
border-radius:
|
| 85 |
-
padding:
|
| 86 |
-
font-size:
|
| 87 |
font-weight: 600 !important;
|
| 88 |
color: white !important;
|
| 89 |
-
|
| 90 |
-
transition: all 0.3s ease !important;
|
| 91 |
}
|
| 92 |
|
| 93 |
.primary-button:hover {
|
| 94 |
-
|
| 95 |
-
|
| 96 |
}
|
| 97 |
|
| 98 |
.secondary-button {
|
| 99 |
-
background:
|
| 100 |
-
border:
|
| 101 |
-
border-radius:
|
| 102 |
-
padding:
|
| 103 |
-
font-size:
|
| 104 |
-
font-weight:
|
| 105 |
color: white !important;
|
| 106 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 107 |
}
|
| 108 |
|
| 109 |
/* Audio Component Styling */
|
| 110 |
.audio-component {
|
| 111 |
-
border-radius:
|
| 112 |
-
border: 2px solid #
|
| 113 |
-
background: #
|
| 114 |
}
|
| 115 |
|
| 116 |
/* Dropdown Styling */
|
| 117 |
.dropdown-component {
|
| 118 |
-
border-radius:
|
| 119 |
-
border: 2px solid #
|
|
|
|
| 120 |
}
|
| 121 |
|
| 122 |
/* Output Text Styling */
|
| 123 |
.output-text {
|
| 124 |
-
background: #
|
| 125 |
-
border-radius:
|
| 126 |
-
border: 2px solid #
|
| 127 |
padding: 20px !important;
|
| 128 |
-
font-size:
|
| 129 |
line-height: 1.6 !important;
|
|
|
|
| 130 |
}
|
| 131 |
|
| 132 |
/* Features Section */
|
| 133 |
.features-section {
|
| 134 |
-
background:
|
| 135 |
-
border
|
|
|
|
| 136 |
padding: 25px;
|
| 137 |
margin-top: 25px;
|
| 138 |
-
box-shadow: 0
|
| 139 |
-
backdrop-filter: blur(10px);
|
| 140 |
-
border: 1px solid rgba(255, 255, 255, 0.2);
|
| 141 |
}
|
| 142 |
|
| 143 |
.feature-item {
|
| 144 |
-
margin-bottom:
|
| 145 |
-
color: #
|
| 146 |
-
font-size:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 147 |
}
|
| 148 |
|
| 149 |
/* Responsive Design */
|
| 150 |
@media (max-width: 768px) {
|
| 151 |
.header-section h1 {
|
| 152 |
-
font-size:
|
| 153 |
}
|
| 154 |
|
| 155 |
.input-card, .output-card {
|
|
@@ -160,20 +171,20 @@ custom_css = """
|
|
| 160 |
"""
|
| 161 |
|
| 162 |
# Create the main interface
|
| 163 |
-
with gr.Blocks(css=custom_css, title="
|
| 164 |
# Header Section
|
| 165 |
with gr.Column(elem_classes="header-section"):
|
| 166 |
gr.HTML("""
|
| 167 |
-
<h1>
|
| 168 |
-
<p>
|
| 169 |
-
<p>
|
| 170 |
""")
|
| 171 |
|
| 172 |
# Main Content
|
| 173 |
with gr.Row():
|
| 174 |
# Input Section
|
| 175 |
with gr.Column(scale=1, elem_classes="input-card"):
|
| 176 |
-
gr.HTML("<h3 style='color: #
|
| 177 |
|
| 178 |
audio_input = gr.Audio(
|
| 179 |
sources=["microphone", "upload"],
|
|
@@ -184,11 +195,11 @@ with gr.Blocks(css=custom_css, title="ποΈ CS-ASR | Code-Switching Speech Rec
|
|
| 184 |
|
| 185 |
model_selector = gr.Dropdown(
|
| 186 |
choices=[
|
| 187 |
-
"
|
| 188 |
-
"
|
| 189 |
],
|
| 190 |
-
value="
|
| 191 |
-
label="
|
| 192 |
elem_classes="dropdown-component",
|
| 193 |
info="Choose the model that best fits your needs"
|
| 194 |
)
|
|
@@ -196,24 +207,24 @@ with gr.Blocks(css=custom_css, title="ποΈ CS-ASR | Code-Switching Speech Rec
|
|
| 196 |
# Action Buttons
|
| 197 |
with gr.Row():
|
| 198 |
transcribe_button = gr.Button(
|
| 199 |
-
"
|
| 200 |
variant="primary",
|
| 201 |
elem_classes="primary-button",
|
| 202 |
scale=2
|
| 203 |
)
|
| 204 |
clear_button = gr.Button(
|
| 205 |
-
"
|
| 206 |
elem_classes="secondary-button",
|
| 207 |
scale=1
|
| 208 |
)
|
| 209 |
|
| 210 |
# Output Section
|
| 211 |
with gr.Column(scale=1, elem_classes="output-card"):
|
| 212 |
-
gr.HTML("<h3 style='color: #
|
| 213 |
|
| 214 |
output_text = gr.Textbox(
|
| 215 |
label="Your Transcription Will Appear Here",
|
| 216 |
-
placeholder="
|
| 217 |
lines=12,
|
| 218 |
elem_classes="output-text",
|
| 219 |
interactive=False
|
|
@@ -222,25 +233,25 @@ with gr.Blocks(css=custom_css, title="ποΈ CS-ASR | Code-Switching Speech Rec
|
|
| 222 |
# Features Section
|
| 223 |
with gr.Column(elem_classes="features-section"):
|
| 224 |
gr.HTML("""
|
| 225 |
-
<h3 style='color: #
|
| 226 |
<div style='display: grid; grid-template-columns: repeat(auto-fit, minmax(250px, 1fr)); gap: 20px; margin-top: 20px;'>
|
| 227 |
<div class='feature-item'>
|
| 228 |
-
<strong>
|
| 229 |
</div>
|
| 230 |
<div class='feature-item'>
|
| 231 |
-
<strong>
|
| 232 |
</div>
|
| 233 |
<div class='feature-item'>
|
| 234 |
-
<strong>
|
| 235 |
</div>
|
| 236 |
<div class='feature-item'>
|
| 237 |
-
<strong>
|
| 238 |
</div>
|
| 239 |
<div class='feature-item'>
|
| 240 |
-
<strong>
|
| 241 |
</div>
|
| 242 |
<div class='feature-item'>
|
| 243 |
-
<strong>
|
| 244 |
</div>
|
| 245 |
</div>
|
| 246 |
""")
|
|
|
|
| 13 |
|
| 14 |
def transcribe(audio, model_choice):
|
| 15 |
if audio is None:
|
| 16 |
+
return "Please upload or record an audio file first."
|
| 17 |
|
| 18 |
try:
|
| 19 |
+
if model_choice == "Whisper (Recommended)":
|
| 20 |
result = whisper_asr(audio)
|
| 21 |
+
return f"Transcription Complete:\n\n{result['text']}"
|
| 22 |
else:
|
| 23 |
result = wav2vec_asr(audio)
|
| 24 |
cleaned_text = clean_transcript(result["text"])
|
| 25 |
+
return f"Transcription Complete:\n\n{cleaned_text}"
|
| 26 |
except Exception as e:
|
| 27 |
+
return f"Error during transcription: {str(e)}"
|
| 28 |
|
| 29 |
def clear_all():
|
| 30 |
+
return None, "Whisper (Recommended)", ""
|
| 31 |
|
| 32 |
+
# Custom CSS for classic professional styling
|
| 33 |
custom_css = """
|
| 34 |
/* Global Styles */
|
| 35 |
.gradio-container {
|
| 36 |
+
background: #f8f9fa !important;
|
| 37 |
+
font-family: 'Times New Roman', Georgia, serif !important;
|
| 38 |
+
color: #2c3e50 !important;
|
| 39 |
}
|
| 40 |
|
| 41 |
/* Header Styling */
|
| 42 |
.header-section {
|
| 43 |
+
background: #ffffff;
|
| 44 |
+
border: 2px solid #dee2e6;
|
| 45 |
+
border-radius: 8px;
|
| 46 |
padding: 30px;
|
| 47 |
margin-bottom: 25px;
|
| 48 |
+
box-shadow: 0 2px 8px rgba(0, 0, 0, 0.1);
|
|
|
|
|
|
|
| 49 |
}
|
| 50 |
|
| 51 |
.header-section h1 {
|
| 52 |
+
color: #2c3e50 !important;
|
| 53 |
+
font-size: 2.2em !important;
|
| 54 |
+
font-weight: 600 !important;
|
|
|
|
|
|
|
| 55 |
text-align: center;
|
| 56 |
margin-bottom: 15px;
|
| 57 |
+
border-bottom: 3px solid #34495e;
|
| 58 |
+
padding-bottom: 15px;
|
| 59 |
}
|
| 60 |
|
| 61 |
.header-section p {
|
| 62 |
+
color: #5a6c7d;
|
| 63 |
font-size: 1.1em;
|
| 64 |
text-align: center;
|
| 65 |
+
margin: 8px 0;
|
| 66 |
+
line-height: 1.5;
|
| 67 |
}
|
| 68 |
|
| 69 |
/* Main Content Cards */
|
| 70 |
.input-card, .output-card {
|
| 71 |
+
background: #ffffff;
|
| 72 |
+
border: 2px solid #dee2e6;
|
| 73 |
+
border-radius: 8px;
|
| 74 |
padding: 25px;
|
| 75 |
+
box-shadow: 0 2px 6px rgba(0, 0, 0, 0.08);
|
|
|
|
|
|
|
| 76 |
margin-bottom: 20px;
|
| 77 |
}
|
| 78 |
|
| 79 |
/* Button Styling */
|
| 80 |
.primary-button {
|
| 81 |
+
background: #34495e !important;
|
| 82 |
+
border: 2px solid #2c3e50 !important;
|
| 83 |
+
border-radius: 6px !important;
|
| 84 |
+
padding: 12px 24px !important;
|
| 85 |
+
font-size: 1em !important;
|
| 86 |
font-weight: 600 !important;
|
| 87 |
color: white !important;
|
| 88 |
+
transition: all 0.2s ease !important;
|
|
|
|
| 89 |
}
|
| 90 |
|
| 91 |
.primary-button:hover {
|
| 92 |
+
background: #2c3e50 !important;
|
| 93 |
+
border-color: #1a252f !important;
|
| 94 |
}
|
| 95 |
|
| 96 |
.secondary-button {
|
| 97 |
+
background: #6c757d !important;
|
| 98 |
+
border: 2px solid #5a6268 !important;
|
| 99 |
+
border-radius: 6px !important;
|
| 100 |
+
padding: 10px 20px !important;
|
| 101 |
+
font-size: 0.95em !important;
|
| 102 |
+
font-weight: 500 !important;
|
| 103 |
color: white !important;
|
| 104 |
+
}
|
| 105 |
+
|
| 106 |
+
.secondary-button:hover {
|
| 107 |
+
background: #5a6268 !important;
|
| 108 |
+
border-color: #495057 !important;
|
| 109 |
}
|
| 110 |
|
| 111 |
/* Audio Component Styling */
|
| 112 |
.audio-component {
|
| 113 |
+
border-radius: 6px !important;
|
| 114 |
+
border: 2px solid #ced4da !important;
|
| 115 |
+
background: #ffffff !important;
|
| 116 |
}
|
| 117 |
|
| 118 |
/* Dropdown Styling */
|
| 119 |
.dropdown-component {
|
| 120 |
+
border-radius: 6px !important;
|
| 121 |
+
border: 2px solid #ced4da !important;
|
| 122 |
+
background: #ffffff !important;
|
| 123 |
}
|
| 124 |
|
| 125 |
/* Output Text Styling */
|
| 126 |
.output-text {
|
| 127 |
+
background: #ffffff !important;
|
| 128 |
+
border-radius: 6px !important;
|
| 129 |
+
border: 2px solid #ced4da !important;
|
| 130 |
padding: 20px !important;
|
| 131 |
+
font-size: 1em !important;
|
| 132 |
line-height: 1.6 !important;
|
| 133 |
+
font-family: 'Courier New', monospace !important;
|
| 134 |
}
|
| 135 |
|
| 136 |
/* Features Section */
|
| 137 |
.features-section {
|
| 138 |
+
background: #ffffff;
|
| 139 |
+
border: 2px solid #dee2e6;
|
| 140 |
+
border-radius: 8px;
|
| 141 |
padding: 25px;
|
| 142 |
margin-top: 25px;
|
| 143 |
+
box-shadow: 0 2px 6px rgba(0, 0, 0, 0.08);
|
|
|
|
|
|
|
| 144 |
}
|
| 145 |
|
| 146 |
.feature-item {
|
| 147 |
+
margin-bottom: 12px;
|
| 148 |
+
color: #495057;
|
| 149 |
+
font-size: 1em;
|
| 150 |
+
line-height: 1.5;
|
| 151 |
+
}
|
| 152 |
+
|
| 153 |
+
/* Professional Typography */
|
| 154 |
+
h3 {
|
| 155 |
+
font-family: 'Times New Roman', Georgia, serif !important;
|
| 156 |
+
color: #2c3e50 !important;
|
| 157 |
+
font-weight: 600 !important;
|
| 158 |
}
|
| 159 |
|
| 160 |
/* Responsive Design */
|
| 161 |
@media (max-width: 768px) {
|
| 162 |
.header-section h1 {
|
| 163 |
+
font-size: 1.8em !important;
|
| 164 |
}
|
| 165 |
|
| 166 |
.input-card, .output-card {
|
|
|
|
| 171 |
"""
|
| 172 |
|
| 173 |
# Create the main interface
|
| 174 |
+
with gr.Blocks(css=custom_css, title="CS-ASR | Code-Switching Speech Recognition") as demo:
|
| 175 |
# Header Section
|
| 176 |
with gr.Column(elem_classes="header-section"):
|
| 177 |
gr.HTML("""
|
| 178 |
+
<h1>Code-Switching ASR Platform</h1>
|
| 179 |
+
<p>Professional Speech Recognition for Khmer-English Code-Switching</p>
|
| 180 |
+
<p>Powered by Advanced Whisper and Wav2Vec2 Models</p>
|
| 181 |
""")
|
| 182 |
|
| 183 |
# Main Content
|
| 184 |
with gr.Row():
|
| 185 |
# Input Section
|
| 186 |
with gr.Column(scale=1, elem_classes="input-card"):
|
| 187 |
+
gr.HTML("<h3 style='color: #2c3e50; margin-bottom: 20px; font-weight: 600;'>Audio Input</h3>")
|
| 188 |
|
| 189 |
audio_input = gr.Audio(
|
| 190 |
sources=["microphone", "upload"],
|
|
|
|
| 195 |
|
| 196 |
model_selector = gr.Dropdown(
|
| 197 |
choices=[
|
| 198 |
+
"Whisper (Recommended)",
|
| 199 |
+
"Wav2Vec2 (Fast)"
|
| 200 |
],
|
| 201 |
+
value="Whisper (Recommended)",
|
| 202 |
+
label="Select AI Model",
|
| 203 |
elem_classes="dropdown-component",
|
| 204 |
info="Choose the model that best fits your needs"
|
| 205 |
)
|
|
|
|
| 207 |
# Action Buttons
|
| 208 |
with gr.Row():
|
| 209 |
transcribe_button = gr.Button(
|
| 210 |
+
"Start Transcription",
|
| 211 |
variant="primary",
|
| 212 |
elem_classes="primary-button",
|
| 213 |
scale=2
|
| 214 |
)
|
| 215 |
clear_button = gr.Button(
|
| 216 |
+
"Clear All",
|
| 217 |
elem_classes="secondary-button",
|
| 218 |
scale=1
|
| 219 |
)
|
| 220 |
|
| 221 |
# Output Section
|
| 222 |
with gr.Column(scale=1, elem_classes="output-card"):
|
| 223 |
+
gr.HTML("<h3 style='color: #2c3e50; margin-bottom: 20px; font-weight: 600;'>Transcription Result</h3>")
|
| 224 |
|
| 225 |
output_text = gr.Textbox(
|
| 226 |
label="Your Transcription Will Appear Here",
|
| 227 |
+
placeholder="Upload an audio file and click 'Start Transcription' to begin processing.",
|
| 228 |
lines=12,
|
| 229 |
elem_classes="output-text",
|
| 230 |
interactive=False
|
|
|
|
| 233 |
# Features Section
|
| 234 |
with gr.Column(elem_classes="features-section"):
|
| 235 |
gr.HTML("""
|
| 236 |
+
<h3 style='color: #2c3e50; margin-bottom: 20px; font-weight: 600; text-align: center;'>Key Features</h3>
|
| 237 |
<div style='display: grid; grid-template-columns: repeat(auto-fit, minmax(250px, 1fr)); gap: 20px; margin-top: 20px;'>
|
| 238 |
<div class='feature-item'>
|
| 239 |
+
<strong>Dual Model Support:</strong> Choose between Whisper and Wav2Vec2 for optimal results
|
| 240 |
</div>
|
| 241 |
<div class='feature-item'>
|
| 242 |
+
<strong>Code-Switching Ready:</strong> Seamlessly handles Khmer-English mixed speech
|
| 243 |
</div>
|
| 244 |
<div class='feature-item'>
|
| 245 |
+
<strong>Flexible Input:</strong> Record live or upload existing audio files
|
| 246 |
</div>
|
| 247 |
<div class='feature-item'>
|
| 248 |
+
<strong>Real-time Processing:</strong> Fast and accurate transcription results
|
| 249 |
</div>
|
| 250 |
<div class='feature-item'>
|
| 251 |
+
<strong>Professional Interface:</strong> Clean, responsive design for all devices
|
| 252 |
</div>
|
| 253 |
<div class='feature-item'>
|
| 254 |
+
<strong>Easy to Use:</strong> No technical knowledge required - just click and transcribe
|
| 255 |
</div>
|
| 256 |
</div>
|
| 257 |
""")
|