Spaces:
Running
Running
Upload app.py
Browse files
app.py
CHANGED
|
@@ -1,56 +1,39 @@
|
|
| 1 |
import gradio as gr
|
| 2 |
import whisper
|
| 3 |
-
import os
|
| 4 |
|
| 5 |
models = {}
|
| 6 |
|
| 7 |
def load_model(model_name):
|
| 8 |
-
global models
|
| 9 |
if model_name not in models:
|
| 10 |
models[model_name] = whisper.load_model(model_name)
|
| 11 |
return models[model_name]
|
| 12 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 13 |
def transcribe(file, model_name, language, show_timestamps, task):
|
| 14 |
if file is None:
|
| 15 |
return "Please upload a video or audio file.", ""
|
| 16 |
-
|
| 17 |
try:
|
| 18 |
m = load_model(model_name)
|
| 19 |
-
|
| 20 |
lang = None if language == "Auto Detect" else language.lower()
|
| 21 |
-
|
| 22 |
-
result = m.transcribe(
|
| 23 |
-
file.name,
|
| 24 |
-
language=lang,
|
| 25 |
-
verbose=False,
|
| 26 |
-
task="translate" if task == "Translate to English" else "transcribe"
|
| 27 |
-
)
|
| 28 |
-
|
| 29 |
-
# Full plain transcript
|
| 30 |
plain = result["text"].strip()
|
| 31 |
-
|
| 32 |
-
# Timestamped transcript
|
| 33 |
if show_timestamps:
|
| 34 |
lines = []
|
| 35 |
for seg in result["segments"]:
|
| 36 |
start = format_time(seg["start"])
|
| 37 |
end = format_time(seg["end"])
|
| 38 |
lines.append(f"[{start} → {end}] {seg['text'].strip()}")
|
| 39 |
-
|
| 40 |
-
|
| 41 |
-
timestamped = plain
|
| 42 |
-
|
| 43 |
-
return timestamped, plain
|
| 44 |
-
|
| 45 |
except Exception as e:
|
| 46 |
return f"Error: {str(e)}", ""
|
| 47 |
|
| 48 |
-
def format_time(seconds):
|
| 49 |
-
m = int(seconds // 60)
|
| 50 |
-
s = int(seconds % 60)
|
| 51 |
-
ms = int((seconds % 1) * 10)
|
| 52 |
-
return f"{m:02d}:{s:02d}.{ms}"
|
| 53 |
-
|
| 54 |
def save_transcript(text):
|
| 55 |
if not text:
|
| 56 |
return None
|
|
@@ -60,22 +43,17 @@ def save_transcript(text):
|
|
| 60 |
return path
|
| 61 |
|
| 62 |
custom_css = """
|
| 63 |
-
@import url('https://fonts.googleapis.com/css2?family=DM+Serif+Display
|
| 64 |
-
|
| 65 |
-
* { box-sizing: border-box; }
|
| 66 |
|
| 67 |
body, .gradio-container {
|
| 68 |
background: #080808 !important;
|
| 69 |
font-family: 'DM Sans', sans-serif !important;
|
| 70 |
}
|
| 71 |
-
|
| 72 |
.gradio-container {
|
| 73 |
max-width: 1000px !important;
|
| 74 |
margin: 0 auto !important;
|
| 75 |
padding: 48px 28px 80px !important;
|
| 76 |
}
|
| 77 |
-
|
| 78 |
-
/* Gold bar */
|
| 79 |
.gradio-container::before {
|
| 80 |
content: '';
|
| 81 |
display: block;
|
|
@@ -84,8 +62,6 @@ body, .gradio-container {
|
|
| 84 |
background: #c9a96e;
|
| 85 |
margin-bottom: 32px;
|
| 86 |
}
|
| 87 |
-
|
| 88 |
-
/* Title */
|
| 89 |
.prose h1 {
|
| 90 |
font-family: 'DM Serif Display', serif !important;
|
| 91 |
font-size: 3rem !important;
|
|
@@ -93,30 +69,19 @@ body, .gradio-container {
|
|
| 93 |
color: #f0ece2 !important;
|
| 94 |
letter-spacing: -0.02em !important;
|
| 95 |
line-height: 1.05 !important;
|
| 96 |
-
margin-bottom: 0 !important;
|
| 97 |
}
|
| 98 |
-
|
| 99 |
.prose p {
|
| 100 |
color: #555 !important;
|
| 101 |
font-size: 0.9rem !important;
|
| 102 |
font-weight: 300 !important;
|
| 103 |
margin-top: 8px !important;
|
| 104 |
-
letter-spacing: 0.01em !important;
|
| 105 |
}
|
| 106 |
-
|
| 107 |
-
/* Panels */
|
| 108 |
.contain, .gap { background: transparent !important; border: none !important; }
|
| 109 |
-
|
| 110 |
.block {
|
| 111 |
background: #0f0f0f !important;
|
| 112 |
border: 1px solid #1c1c1c !important;
|
| 113 |
border-radius: 14px !important;
|
| 114 |
-
transition: border-color 0.2s !important;
|
| 115 |
}
|
| 116 |
-
|
| 117 |
-
.block:hover { border-color: #2a2a2a !important; }
|
| 118 |
-
|
| 119 |
-
/* Labels */
|
| 120 |
.block label > span, label > span {
|
| 121 |
font-family: 'DM Sans', sans-serif !important;
|
| 122 |
font-size: 0.72rem !important;
|
|
@@ -125,18 +90,6 @@ body, .gradio-container {
|
|
| 125 |
text-transform: uppercase !important;
|
| 126 |
letter-spacing: 0.12em !important;
|
| 127 |
}
|
| 128 |
-
|
| 129 |
-
/* Dropdowns & selects */
|
| 130 |
-
select, .wrap-inner, input[type="text"] {
|
| 131 |
-
background: #0f0f0f !important;
|
| 132 |
-
color: #c8c4bc !important;
|
| 133 |
-
border: 1px solid #1c1c1c !important;
|
| 134 |
-
border-radius: 10px !important;
|
| 135 |
-
font-family: 'DM Sans', sans-serif !important;
|
| 136 |
-
font-size: 0.88rem !important;
|
| 137 |
-
}
|
| 138 |
-
|
| 139 |
-
/* Textarea */
|
| 140 |
textarea {
|
| 141 |
background: transparent !important;
|
| 142 |
color: #d4d0c8 !important;
|
|
@@ -145,15 +98,9 @@ textarea {
|
|
| 145 |
line-height: 1.85 !important;
|
| 146 |
font-weight: 300 !important;
|
| 147 |
border: none !important;
|
| 148 |
-
letter-spacing: 0.01em !important;
|
| 149 |
}
|
| 150 |
-
|
| 151 |
textarea::placeholder { color: #2e2e2e !important; }
|
| 152 |
-
|
| 153 |
-
/* Checkbox */
|
| 154 |
input[type="checkbox"] { accent-color: #c9a96e !important; }
|
| 155 |
-
|
| 156 |
-
/* Primary button */
|
| 157 |
button.primary {
|
| 158 |
background: #c9a96e !important;
|
| 159 |
border: none !important;
|
|
@@ -166,18 +113,12 @@ button.primary {
|
|
| 166 |
text-transform: uppercase !important;
|
| 167 |
padding: 14px 28px !important;
|
| 168 |
transition: all 0.2s ease !important;
|
| 169 |
-
box-shadow: 0 2px 12px rgba(201,169,110,0.15) !important;
|
| 170 |
}
|
| 171 |
-
|
| 172 |
button.primary:hover {
|
| 173 |
background: #d4b87e !important;
|
| 174 |
transform: translateY(-1px) !important;
|
| 175 |
box-shadow: 0 6px 24px rgba(201,169,110,0.3) !important;
|
| 176 |
}
|
| 177 |
-
|
| 178 |
-
button.primary:active { transform: translateY(0) !important; }
|
| 179 |
-
|
| 180 |
-
/* Secondary button */
|
| 181 |
button.secondary {
|
| 182 |
background: transparent !important;
|
| 183 |
border: 1px solid #1e1e1e !important;
|
|
@@ -190,19 +131,7 @@ button.secondary {
|
|
| 190 |
text-transform: uppercase !important;
|
| 191 |
transition: all 0.2s ease !important;
|
| 192 |
}
|
| 193 |
-
|
| 194 |
-
button.secondary:hover {
|
| 195 |
-
border-color: #3a3a3a !important;
|
| 196 |
-
color: #c8c4bc !important;
|
| 197 |
-
}
|
| 198 |
-
|
| 199 |
-
/* Download button area */
|
| 200 |
-
.file-preview {
|
| 201 |
-
background: transparent !important;
|
| 202 |
-
border: none !important;
|
| 203 |
-
}
|
| 204 |
-
|
| 205 |
-
/* Tabs */
|
| 206 |
.tab-nav button {
|
| 207 |
font-family: 'DM Sans', sans-serif !important;
|
| 208 |
font-size: 0.78rem !important;
|
|
@@ -213,31 +142,13 @@ button.secondary:hover {
|
|
| 213 |
background: transparent !important;
|
| 214 |
border: none !important;
|
| 215 |
border-bottom: 2px solid transparent !important;
|
| 216 |
-
padding: 10px 16px !important;
|
| 217 |
transition: all 0.2s !important;
|
| 218 |
}
|
| 219 |
-
|
| 220 |
-
.tab-nav button.selected {
|
| 221 |
-
color: #c9a96e !important;
|
| 222 |
-
border-bottom-color: #c9a96e !important;
|
| 223 |
-
}
|
| 224 |
-
|
| 225 |
-
/* Divider */
|
| 226 |
-
.divider { border-color: #1c1c1c !important; }
|
| 227 |
-
|
| 228 |
-
/* Progress */
|
| 229 |
-
.progress-bar { background: #c9a96e !important; }
|
| 230 |
-
.progress-bar-wrap { background: #1c1c1c !important; }
|
| 231 |
-
|
| 232 |
-
/* Scrollbar */
|
| 233 |
::-webkit-scrollbar { width: 3px; }
|
| 234 |
::-webkit-scrollbar-track { background: transparent; }
|
| 235 |
::-webkit-scrollbar-thumb { background: #2a2a2a; border-radius: 4px; }
|
| 236 |
-
|
| 237 |
footer { display: none !important; }
|
| 238 |
-
|
| 239 |
-
/* Settings row */
|
| 240 |
-
.settings-row { gap: 12px !important; }
|
| 241 |
"""
|
| 242 |
|
| 243 |
LANGUAGES = [
|
|
@@ -246,80 +157,54 @@ LANGUAGES = [
|
|
| 246 |
"Korean", "Arabic", "Russian", "Dutch", "Turkish"
|
| 247 |
]
|
| 248 |
|
| 249 |
-
with gr.Blocks(
|
| 250 |
-
|
| 251 |
gr.Markdown("""
|
| 252 |
# KalpTranscript
|
| 253 |
Turn any video or audio into clean, readable text — powered by OpenAI Whisper.
|
| 254 |
""")
|
| 255 |
|
| 256 |
with gr.Row():
|
| 257 |
-
# Left column - controls
|
| 258 |
with gr.Column(scale=1):
|
| 259 |
file_input = gr.File(
|
| 260 |
label="Drop your file here",
|
| 261 |
file_types=["video", "audio"]
|
| 262 |
)
|
| 263 |
-
|
| 264 |
-
with gr.Row(elem_classes=["settings-row"]):
|
| 265 |
model_choice = gr.Dropdown(
|
| 266 |
choices=["tiny", "base", "small", "medium"],
|
| 267 |
value="small",
|
| 268 |
-
label="Model"
|
| 269 |
-
info="small = fast & accurate, medium = best quality"
|
| 270 |
)
|
| 271 |
language = gr.Dropdown(
|
| 272 |
choices=LANGUAGES,
|
| 273 |
value="Auto Detect",
|
| 274 |
label="Language"
|
| 275 |
)
|
| 276 |
-
|
| 277 |
-
|
| 278 |
-
|
| 279 |
-
|
| 280 |
-
value="Transcribe",
|
| 281 |
-
label="Task"
|
| 282 |
-
)
|
| 283 |
-
|
| 284 |
-
timestamps = gr.Checkbox(
|
| 285 |
-
label="Show timestamps",
|
| 286 |
-
value=False
|
| 287 |
)
|
| 288 |
-
|
| 289 |
with gr.Row():
|
| 290 |
clear_btn = gr.ClearButton(value="Clear")
|
| 291 |
submit_btn = gr.Button("Transcribe →", variant="primary")
|
| 292 |
|
| 293 |
-
# Right column - output
|
| 294 |
with gr.Column(scale=1):
|
| 295 |
with gr.Tabs():
|
| 296 |
with gr.Tab("Transcript"):
|
| 297 |
-
output = gr.Textbox(
|
| 298 |
-
label="",
|
| 299 |
-
lines=16,
|
| 300 |
-
placeholder="Your transcript will appear here...",
|
| 301 |
-
show_copy_button=False
|
| 302 |
-
)
|
| 303 |
with gr.Tab("Download"):
|
| 304 |
-
plain_output = gr.Textbox(
|
| 305 |
-
label="Plain text (for download)",
|
| 306 |
-
lines=10,
|
| 307 |
-
visible=False
|
| 308 |
-
)
|
| 309 |
download_btn = gr.Button("Save as .txt", variant="secondary")
|
| 310 |
-
download_file = gr.File(label="Download"
|
| 311 |
|
| 312 |
submit_btn.click(
|
| 313 |
fn=transcribe,
|
| 314 |
inputs=[file_input, model_choice, language, timestamps, task],
|
| 315 |
outputs=[output, plain_output]
|
| 316 |
)
|
| 317 |
-
|
| 318 |
-
download_btn.click(
|
| 319 |
-
fn=save_transcript,
|
| 320 |
-
inputs=plain_output,
|
| 321 |
-
outputs=download_file
|
| 322 |
-
)
|
| 323 |
|
| 324 |
if __name__ == "__main__":
|
| 325 |
-
demo.launch()
|
|
|
|
| 1 |
import gradio as gr
|
| 2 |
import whisper
|
|
|
|
| 3 |
|
| 4 |
models = {}
|
| 5 |
|
| 6 |
def load_model(model_name):
|
|
|
|
| 7 |
if model_name not in models:
|
| 8 |
models[model_name] = whisper.load_model(model_name)
|
| 9 |
return models[model_name]
|
| 10 |
|
| 11 |
+
def format_time(seconds):
|
| 12 |
+
m = int(seconds // 60)
|
| 13 |
+
s = int(seconds % 60)
|
| 14 |
+
ms = int((seconds % 1) * 10)
|
| 15 |
+
return f"{m:02d}:{s:02d}.{ms}"
|
| 16 |
+
|
| 17 |
def transcribe(file, model_name, language, show_timestamps, task):
|
| 18 |
if file is None:
|
| 19 |
return "Please upload a video or audio file.", ""
|
|
|
|
| 20 |
try:
|
| 21 |
m = load_model(model_name)
|
|
|
|
| 22 |
lang = None if language == "Auto Detect" else language.lower()
|
| 23 |
+
whisper_task = "translate" if task == "Translate to English" else "transcribe"
|
| 24 |
+
result = m.transcribe(file.name, language=lang, verbose=False, task=whisper_task)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 25 |
plain = result["text"].strip()
|
|
|
|
|
|
|
| 26 |
if show_timestamps:
|
| 27 |
lines = []
|
| 28 |
for seg in result["segments"]:
|
| 29 |
start = format_time(seg["start"])
|
| 30 |
end = format_time(seg["end"])
|
| 31 |
lines.append(f"[{start} → {end}] {seg['text'].strip()}")
|
| 32 |
+
return "\n".join(lines), plain
|
| 33 |
+
return plain, plain
|
|
|
|
|
|
|
|
|
|
|
|
|
| 34 |
except Exception as e:
|
| 35 |
return f"Error: {str(e)}", ""
|
| 36 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 37 |
def save_transcript(text):
|
| 38 |
if not text:
|
| 39 |
return None
|
|
|
|
| 43 |
return path
|
| 44 |
|
| 45 |
custom_css = """
|
| 46 |
+
@import url('https://fonts.googleapis.com/css2?family=DM+Serif+Display&family=DM+Sans:wght@300;400;500&display=swap');
|
|
|
|
|
|
|
| 47 |
|
| 48 |
body, .gradio-container {
|
| 49 |
background: #080808 !important;
|
| 50 |
font-family: 'DM Sans', sans-serif !important;
|
| 51 |
}
|
|
|
|
| 52 |
.gradio-container {
|
| 53 |
max-width: 1000px !important;
|
| 54 |
margin: 0 auto !important;
|
| 55 |
padding: 48px 28px 80px !important;
|
| 56 |
}
|
|
|
|
|
|
|
| 57 |
.gradio-container::before {
|
| 58 |
content: '';
|
| 59 |
display: block;
|
|
|
|
| 62 |
background: #c9a96e;
|
| 63 |
margin-bottom: 32px;
|
| 64 |
}
|
|
|
|
|
|
|
| 65 |
.prose h1 {
|
| 66 |
font-family: 'DM Serif Display', serif !important;
|
| 67 |
font-size: 3rem !important;
|
|
|
|
| 69 |
color: #f0ece2 !important;
|
| 70 |
letter-spacing: -0.02em !important;
|
| 71 |
line-height: 1.05 !important;
|
|
|
|
| 72 |
}
|
|
|
|
| 73 |
.prose p {
|
| 74 |
color: #555 !important;
|
| 75 |
font-size: 0.9rem !important;
|
| 76 |
font-weight: 300 !important;
|
| 77 |
margin-top: 8px !important;
|
|
|
|
| 78 |
}
|
|
|
|
|
|
|
| 79 |
.contain, .gap { background: transparent !important; border: none !important; }
|
|
|
|
| 80 |
.block {
|
| 81 |
background: #0f0f0f !important;
|
| 82 |
border: 1px solid #1c1c1c !important;
|
| 83 |
border-radius: 14px !important;
|
|
|
|
| 84 |
}
|
|
|
|
|
|
|
|
|
|
|
|
|
| 85 |
.block label > span, label > span {
|
| 86 |
font-family: 'DM Sans', sans-serif !important;
|
| 87 |
font-size: 0.72rem !important;
|
|
|
|
| 90 |
text-transform: uppercase !important;
|
| 91 |
letter-spacing: 0.12em !important;
|
| 92 |
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 93 |
textarea {
|
| 94 |
background: transparent !important;
|
| 95 |
color: #d4d0c8 !important;
|
|
|
|
| 98 |
line-height: 1.85 !important;
|
| 99 |
font-weight: 300 !important;
|
| 100 |
border: none !important;
|
|
|
|
| 101 |
}
|
|
|
|
| 102 |
textarea::placeholder { color: #2e2e2e !important; }
|
|
|
|
|
|
|
| 103 |
input[type="checkbox"] { accent-color: #c9a96e !important; }
|
|
|
|
|
|
|
| 104 |
button.primary {
|
| 105 |
background: #c9a96e !important;
|
| 106 |
border: none !important;
|
|
|
|
| 113 |
text-transform: uppercase !important;
|
| 114 |
padding: 14px 28px !important;
|
| 115 |
transition: all 0.2s ease !important;
|
|
|
|
| 116 |
}
|
|
|
|
| 117 |
button.primary:hover {
|
| 118 |
background: #d4b87e !important;
|
| 119 |
transform: translateY(-1px) !important;
|
| 120 |
box-shadow: 0 6px 24px rgba(201,169,110,0.3) !important;
|
| 121 |
}
|
|
|
|
|
|
|
|
|
|
|
|
|
| 122 |
button.secondary {
|
| 123 |
background: transparent !important;
|
| 124 |
border: 1px solid #1e1e1e !important;
|
|
|
|
| 131 |
text-transform: uppercase !important;
|
| 132 |
transition: all 0.2s ease !important;
|
| 133 |
}
|
| 134 |
+
button.secondary:hover { border-color: #3a3a3a !important; color: #c8c4bc !important; }
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 135 |
.tab-nav button {
|
| 136 |
font-family: 'DM Sans', sans-serif !important;
|
| 137 |
font-size: 0.78rem !important;
|
|
|
|
| 142 |
background: transparent !important;
|
| 143 |
border: none !important;
|
| 144 |
border-bottom: 2px solid transparent !important;
|
|
|
|
| 145 |
transition: all 0.2s !important;
|
| 146 |
}
|
| 147 |
+
.tab-nav button.selected { color: #c9a96e !important; border-bottom-color: #c9a96e !important; }
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 148 |
::-webkit-scrollbar { width: 3px; }
|
| 149 |
::-webkit-scrollbar-track { background: transparent; }
|
| 150 |
::-webkit-scrollbar-thumb { background: #2a2a2a; border-radius: 4px; }
|
|
|
|
| 151 |
footer { display: none !important; }
|
|
|
|
|
|
|
|
|
|
| 152 |
"""
|
| 153 |
|
| 154 |
LANGUAGES = [
|
|
|
|
| 157 |
"Korean", "Arabic", "Russian", "Dutch", "Turkish"
|
| 158 |
]
|
| 159 |
|
| 160 |
+
with gr.Blocks(title="KalpTranscript") as demo:
|
|
|
|
| 161 |
gr.Markdown("""
|
| 162 |
# KalpTranscript
|
| 163 |
Turn any video or audio into clean, readable text — powered by OpenAI Whisper.
|
| 164 |
""")
|
| 165 |
|
| 166 |
with gr.Row():
|
|
|
|
| 167 |
with gr.Column(scale=1):
|
| 168 |
file_input = gr.File(
|
| 169 |
label="Drop your file here",
|
| 170 |
file_types=["video", "audio"]
|
| 171 |
)
|
| 172 |
+
with gr.Row():
|
|
|
|
| 173 |
model_choice = gr.Dropdown(
|
| 174 |
choices=["tiny", "base", "small", "medium"],
|
| 175 |
value="small",
|
| 176 |
+
label="Model"
|
|
|
|
| 177 |
)
|
| 178 |
language = gr.Dropdown(
|
| 179 |
choices=LANGUAGES,
|
| 180 |
value="Auto Detect",
|
| 181 |
label="Language"
|
| 182 |
)
|
| 183 |
+
task = gr.Radio(
|
| 184 |
+
choices=["Transcribe", "Translate to English"],
|
| 185 |
+
value="Transcribe",
|
| 186 |
+
label="Task"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 187 |
)
|
| 188 |
+
timestamps = gr.Checkbox(label="Show timestamps", value=False)
|
| 189 |
with gr.Row():
|
| 190 |
clear_btn = gr.ClearButton(value="Clear")
|
| 191 |
submit_btn = gr.Button("Transcribe →", variant="primary")
|
| 192 |
|
|
|
|
| 193 |
with gr.Column(scale=1):
|
| 194 |
with gr.Tabs():
|
| 195 |
with gr.Tab("Transcript"):
|
| 196 |
+
output = gr.Textbox(label="", lines=16, placeholder="Your transcript will appear here...")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 197 |
with gr.Tab("Download"):
|
| 198 |
+
plain_output = gr.Textbox(label="Plain text", lines=10, visible=False)
|
|
|
|
|
|
|
|
|
|
|
|
|
| 199 |
download_btn = gr.Button("Save as .txt", variant="secondary")
|
| 200 |
+
download_file = gr.File(label="Download")
|
| 201 |
|
| 202 |
submit_btn.click(
|
| 203 |
fn=transcribe,
|
| 204 |
inputs=[file_input, model_choice, language, timestamps, task],
|
| 205 |
outputs=[output, plain_output]
|
| 206 |
)
|
| 207 |
+
download_btn.click(fn=save_transcript, inputs=plain_output, outputs=download_file)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 208 |
|
| 209 |
if __name__ == "__main__":
|
| 210 |
+
demo.launch(css=custom_css)
|