Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
|
@@ -139,48 +139,23 @@ class MultiASRApp:
|
|
| 139 |
# ------------------------
|
| 140 |
# Transcription
|
| 141 |
# ------------------------
|
| 142 |
-
def transcribe(self, audio, model_name
|
| 143 |
if audio is None:
|
| 144 |
-
return "Ανέβασε ένα ηχητικό αρχείο.", ""
|
| 145 |
|
| 146 |
start = time.time()
|
| 147 |
if not self.load_model(model_name):
|
| 148 |
-
return "Σφάλμα φόρτωσης μοντέλου.", ""
|
| 149 |
|
| 150 |
-
|
| 151 |
-
|
| 152 |
-
text = result.get("text", "")
|
| 153 |
|
| 154 |
-
|
| 155 |
-
"Οι χρονικές σημάνσεις δεν υποστηρίζονται για αυτό το μοντέλο."
|
| 156 |
-
if return_timestamps
|
| 157 |
-
else ""
|
| 158 |
-
)
|
| 159 |
-
|
| 160 |
-
else:
|
| 161 |
-
result = self.pipe(
|
| 162 |
-
audio,
|
| 163 |
-
return_timestamps=return_timestamps,
|
| 164 |
-
)
|
| 165 |
-
text = result.get("text", "")
|
| 166 |
-
timestamps = self.format_timestamps(result.get("chunks", []))
|
| 167 |
-
|
| 168 |
-
details = (
|
| 169 |
f"Μοντέλο: {model_name}\n"
|
| 170 |
f"Χρόνος επεξεργασίας: {time.time() - start:.2f} δευτ."
|
| 171 |
)
|
| 172 |
|
| 173 |
-
return text.strip(),
|
| 174 |
-
|
| 175 |
-
def format_timestamps(self, chunks):
|
| 176 |
-
if not chunks:
|
| 177 |
-
return ""
|
| 178 |
-
out = ""
|
| 179 |
-
for c in chunks:
|
| 180 |
-
ts = c.get("timestamp")
|
| 181 |
-
if ts and ts[0] is not None and ts[1] is not None:
|
| 182 |
-
out += f"[{ts[0]:.1f}–{ts[1]:.1f}] {c.get('text','')}\n"
|
| 183 |
-
return out
|
| 184 |
|
| 185 |
def status(self):
|
| 186 |
if not self.current_model:
|
|
@@ -193,8 +168,8 @@ class MultiASRApp:
|
|
| 193 |
# ------------------------
|
| 194 |
app = MultiASRApp()
|
| 195 |
|
| 196 |
-
def run(audio, model
|
| 197 |
-
return app.transcribe(audio, model
|
| 198 |
|
| 199 |
def status():
|
| 200 |
return app.status()
|
|
@@ -214,11 +189,18 @@ with gr.Blocks(title="Ίντα λαλείς;", theme=gr.themes.Soft()) as demo:
|
|
| 214 |
"""
|
| 215 |
)
|
| 216 |
|
| 217 |
-
model_status = gr.Textbox(
|
|
|
|
|
|
|
|
|
|
|
|
|
| 218 |
|
| 219 |
with gr.Row():
|
| 220 |
with gr.Column():
|
| 221 |
-
audio = gr.Audio(
|
|
|
|
|
|
|
|
|
|
| 222 |
|
| 223 |
model = gr.Dropdown(
|
| 224 |
choices=app.available_models,
|
|
@@ -226,19 +208,27 @@ with gr.Blocks(title="Ίντα λαλείς;", theme=gr.themes.Soft()) as demo:
|
|
| 226 |
label="Μοντέλο αναγνώρισης ομιλίας",
|
| 227 |
)
|
| 228 |
|
| 229 |
-
|
| 230 |
-
|
| 231 |
-
|
|
|
|
| 232 |
|
| 233 |
with gr.Column():
|
| 234 |
-
text_out = gr.Textbox(
|
| 235 |
-
|
| 236 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 237 |
|
| 238 |
btn.click(
|
| 239 |
run,
|
| 240 |
-
inputs=[audio, model
|
| 241 |
-
outputs=[text_out,
|
| 242 |
)
|
| 243 |
|
| 244 |
model.change(lambda _: status(), outputs=model_status)
|
|
|
|
| 139 |
# ------------------------
|
| 140 |
# Transcription
|
| 141 |
# ------------------------
|
| 142 |
+
def transcribe(self, audio, model_name):
|
| 143 |
if audio is None:
|
| 144 |
+
return "Ανέβασε ένα ηχητικό αρχείο.", ""
|
| 145 |
|
| 146 |
start = time.time()
|
| 147 |
if not self.load_model(model_name):
|
| 148 |
+
return "Σφάλμα φόρτωσης μοντέλου.", ""
|
| 149 |
|
| 150 |
+
result = self.pipe(audio)
|
| 151 |
+
text = result.get("text", "")
|
|
|
|
| 152 |
|
| 153 |
+
info = (
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 154 |
f"Μοντέλο: {model_name}\n"
|
| 155 |
f"Χρόνος επεξεργασίας: {time.time() - start:.2f} δευτ."
|
| 156 |
)
|
| 157 |
|
| 158 |
+
return text.strip(), info
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 159 |
|
| 160 |
def status(self):
|
| 161 |
if not self.current_model:
|
|
|
|
| 168 |
# ------------------------
|
| 169 |
app = MultiASRApp()
|
| 170 |
|
| 171 |
+
def run(audio, model):
|
| 172 |
+
return app.transcribe(audio, model)
|
| 173 |
|
| 174 |
def status():
|
| 175 |
return app.status()
|
|
|
|
| 189 |
"""
|
| 190 |
)
|
| 191 |
|
| 192 |
+
model_status = gr.Textbox(
|
| 193 |
+
label="Κατάσταση μοντέλου",
|
| 194 |
+
value=status(),
|
| 195 |
+
interactive=False,
|
| 196 |
+
)
|
| 197 |
|
| 198 |
with gr.Row():
|
| 199 |
with gr.Column():
|
| 200 |
+
audio = gr.Audio(
|
| 201 |
+
label="🎵 Ανέβασε ηχητικό αρχείο",
|
| 202 |
+
type="filepath",
|
| 203 |
+
)
|
| 204 |
|
| 205 |
model = gr.Dropdown(
|
| 206 |
choices=app.available_models,
|
|
|
|
| 208 |
label="Μοντέλο αναγνώρισης ομιλίας",
|
| 209 |
)
|
| 210 |
|
| 211 |
+
btn = gr.Button(
|
| 212 |
+
"🗣️ Μετατροπή ομιλίας σε κείμενο",
|
| 213 |
+
variant="primary",
|
| 214 |
+
)
|
| 215 |
|
| 216 |
with gr.Column():
|
| 217 |
+
text_out = gr.Textbox(
|
| 218 |
+
label="📄 Κείμενο",
|
| 219 |
+
lines=8,
|
| 220 |
+
show_copy_button=True,
|
| 221 |
+
)
|
| 222 |
+
|
| 223 |
+
info_out = gr.Textbox(
|
| 224 |
+
label="Πληροφορίες",
|
| 225 |
+
lines=4,
|
| 226 |
+
)
|
| 227 |
|
| 228 |
btn.click(
|
| 229 |
run,
|
| 230 |
+
inputs=[audio, model],
|
| 231 |
+
outputs=[text_out, info_out],
|
| 232 |
)
|
| 233 |
|
| 234 |
model.change(lambda _: status(), outputs=model_status)
|