Spaces:
Running
on
Zero
Running
on
Zero
Nithin Rao Koluguri
commited on
Commit
·
f4154c5
1
Parent(s):
0e3aa4b
Add SRT download button
Browse filesSigned-off-by: Nithin Rao Koluguri <nithinraok>
app.py
CHANGED
@@ -10,6 +10,7 @@ import numpy as np
|
|
10 |
import os
|
11 |
import gradio.themes as gr_themes
|
12 |
import csv
|
|
|
13 |
|
14 |
device = "cuda" if torch.cuda.is_available() else "cpu"
|
15 |
MODEL_NAME="nvidia/parakeet-tdt-0.6b-v2"
|
@@ -72,20 +73,52 @@ def get_audio_segment(audio_path, start_second, end_second):
|
|
72 |
print(f"Error clipping audio {audio_path} from {start_second}s to {end_second}s: {e}")
|
73 |
return None
|
74 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
75 |
@spaces.GPU
|
76 |
def get_transcripts_and_raw_times(audio_path, session_dir):
|
77 |
if not audio_path:
|
78 |
gr.Error("No audio file path provided for transcription.", duration=None)
|
79 |
-
# Return an update to hide the
|
80 |
-
return [], [], None, gr.DownloadButton(visible=False)
|
81 |
|
82 |
vis_data = [["N/A", "N/A", "Processing failed"]]
|
83 |
raw_times_data = [[0.0, 0.0]]
|
84 |
processed_audio_path = None
|
85 |
csv_file_path = None
|
|
|
86 |
original_path_name = Path(audio_path).name
|
87 |
audio_name = Path(audio_path).stem
|
88 |
|
|
|
|
|
|
|
|
|
89 |
try:
|
90 |
try:
|
91 |
gr.Info(f"Loading audio: {original_path_name}", duration=2)
|
@@ -93,8 +126,7 @@ def get_transcripts_and_raw_times(audio_path, session_dir):
|
|
93 |
duration_sec = audio.duration_seconds
|
94 |
except Exception as load_e:
|
95 |
gr.Error(f"Failed to load audio file {original_path_name}: {load_e}", duration=None)
|
96 |
-
|
97 |
-
return [["Error", "Error", "Load failed"]], [[0.0, 0.0]], audio_path, gr.DownloadButton(visible=False)
|
98 |
|
99 |
resampled = False
|
100 |
mono = False
|
@@ -106,8 +138,7 @@ def get_transcripts_and_raw_times(audio_path, session_dir):
|
|
106 |
resampled = True
|
107 |
except Exception as resample_e:
|
108 |
gr.Error(f"Failed to resample audio: {resample_e}", duration=None)
|
109 |
-
|
110 |
-
return [["Error", "Error", "Resample failed"]], [[0.0, 0.0]], audio_path, gr.DownloadButton(visible=False)
|
111 |
|
112 |
if audio.channels == 2:
|
113 |
try:
|
@@ -115,12 +146,10 @@ def get_transcripts_and_raw_times(audio_path, session_dir):
|
|
115 |
mono = True
|
116 |
except Exception as mono_e:
|
117 |
gr.Error(f"Failed to convert audio to mono: {mono_e}", duration=None)
|
118 |
-
|
119 |
-
return [["Error", "Error", "Mono conversion failed"]], [[0.0, 0.0]], audio_path, gr.DownloadButton(visible=False)
|
120 |
elif audio.channels > 2:
|
121 |
gr.Error(f"Audio has {audio.channels} channels. Only mono (1) or stereo (2) supported.", duration=None)
|
122 |
-
|
123 |
-
return [["Error", "Error", f"{audio.channels}-channel audio not supported"]], [[0.0, 0.0]], audio_path, gr.DownloadButton(visible=False)
|
124 |
|
125 |
if resampled or mono:
|
126 |
try:
|
@@ -132,8 +161,7 @@ def get_transcripts_and_raw_times(audio_path, session_dir):
|
|
132 |
gr.Error(f"Failed to export processed audio: {export_e}", duration=None)
|
133 |
if processed_audio_path and os.path.exists(processed_audio_path):
|
134 |
os.remove(processed_audio_path)
|
135 |
-
|
136 |
-
return [["Error", "Error", "Export failed"]], [[0.0, 0.0]], audio_path, gr.DownloadButton(visible=False)
|
137 |
else:
|
138 |
transcribe_path = audio_path
|
139 |
info_path_name = original_path_name
|
@@ -163,46 +191,52 @@ def get_transcripts_and_raw_times(audio_path, session_dir):
|
|
163 |
|
164 |
if not output or not isinstance(output, list) or not output[0] or not hasattr(output[0], 'timestamp') or not output[0].timestamp or 'segment' not in output[0].timestamp:
|
165 |
gr.Error("Transcription failed or produced unexpected output format.", duration=None)
|
166 |
-
# Return an update to hide the
|
167 |
-
return [["Error", "Error", "Transcription Format Issue"]], [[0.0, 0.0]], audio_path,
|
168 |
|
169 |
segment_timestamps = output[0].timestamp['segment']
|
170 |
csv_headers = ["Start (s)", "End (s)", "Segment"]
|
171 |
vis_data = [[f"{ts['start']:.2f}", f"{ts['end']:.2f}", ts['segment']] for ts in segment_timestamps]
|
172 |
raw_times_data = [[ts['start'], ts['end']] for ts in segment_timestamps]
|
173 |
|
174 |
-
#
|
175 |
-
button_update = gr.DownloadButton(visible=False)
|
176 |
try:
|
177 |
csv_file_path = Path(session_dir, f"transcription_{audio_name}.csv")
|
178 |
writer = csv.writer(open(csv_file_path, 'w'))
|
179 |
writer.writerow(csv_headers)
|
180 |
writer.writerows(vis_data)
|
181 |
print(f"CSV transcript saved to temporary file: {csv_file_path}")
|
182 |
-
|
183 |
-
button_update = gr.DownloadButton(value=csv_file_path, visible=True)
|
184 |
except Exception as csv_e:
|
185 |
gr.Error(f"Failed to create transcript CSV file: {csv_e}", duration=None)
|
186 |
print(f"Error writing CSV: {csv_e}")
|
187 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
188 |
|
189 |
gr.Info("Transcription complete.", duration=2)
|
190 |
-
|
191 |
-
return vis_data, raw_times_data, audio_path, button_update
|
192 |
|
193 |
except torch.cuda.OutOfMemoryError as e:
|
194 |
error_msg = 'CUDA out of memory. Please try a shorter audio or reduce GPU load.'
|
195 |
print(f"CUDA OutOfMemoryError: {e}")
|
196 |
gr.Error(error_msg, duration=None)
|
197 |
-
|
198 |
-
return [["OOM", "OOM", error_msg]], [[0.0, 0.0]], audio_path, gr.DownloadButton(visible=False)
|
199 |
|
200 |
except FileNotFoundError:
|
201 |
error_msg = f"Audio file for transcription not found: {Path(transcribe_path).name}."
|
202 |
print(f"Error: Transcribe audio file not found at path: {transcribe_path}")
|
203 |
gr.Error(error_msg, duration=None)
|
204 |
-
|
205 |
-
return [["Error", "Error", "File not found for transcription"]], [[0.0, 0.0]], audio_path, gr.DownloadButton(visible=False)
|
206 |
|
207 |
except Exception as e:
|
208 |
error_msg = f"Transcription failed: {e}"
|
@@ -210,8 +244,7 @@ def get_transcripts_and_raw_times(audio_path, session_dir):
|
|
210 |
gr.Error(error_msg, duration=None)
|
211 |
vis_data = [["Error", "Error", error_msg]]
|
212 |
raw_times_data = [[0.0, 0.0]]
|
213 |
-
|
214 |
-
return vis_data, raw_times_data, audio_path, gr.DownloadButton(visible=False)
|
215 |
finally:
|
216 |
# --- Model Cleanup ---
|
217 |
try:
|
@@ -349,7 +382,9 @@ with gr.Blocks(theme=nvidia_theme) as demo:
|
|
349 |
gr.Markdown("<p><strong style='color: #FF0000; font-size: 1.2em;'>Transcription Results (Click row to play segment)</strong></p>")
|
350 |
|
351 |
# Define the DownloadButton *before* the DataFrame
|
352 |
-
|
|
|
|
|
353 |
|
354 |
vis_timestamps_df = gr.DataFrame(
|
355 |
headers=["Start (s)", "End (s)", "Segment"],
|
@@ -364,14 +399,14 @@ with gr.Blocks(theme=nvidia_theme) as demo:
|
|
364 |
mic_transcribe_btn.click(
|
365 |
fn=get_transcripts_and_raw_times,
|
366 |
inputs=[mic_input, session_dir],
|
367 |
-
outputs=[vis_timestamps_df, raw_timestamps_list_state, current_audio_path_state,
|
368 |
api_name="transcribe_mic"
|
369 |
)
|
370 |
|
371 |
file_transcribe_btn.click(
|
372 |
fn=get_transcripts_and_raw_times,
|
373 |
inputs=[file_input, session_dir],
|
374 |
-
outputs=[vis_timestamps_df, raw_timestamps_list_state, current_audio_path_state,
|
375 |
api_name="transcribe_file"
|
376 |
)
|
377 |
|
|
|
10 |
import os
|
11 |
import gradio.themes as gr_themes
|
12 |
import csv
|
13 |
+
import datetime
|
14 |
|
15 |
device = "cuda" if torch.cuda.is_available() else "cpu"
|
16 |
MODEL_NAME="nvidia/parakeet-tdt-0.6b-v2"
|
|
|
73 |
print(f"Error clipping audio {audio_path} from {start_second}s to {end_second}s: {e}")
|
74 |
return None
|
75 |
|
76 |
+
def format_srt_time(seconds: float) -> str:
|
77 |
+
"""Converts seconds to SRT time format HH:MM:SS,mmm using datetime.timedelta"""
|
78 |
+
sanitized_total_seconds = max(0.0, seconds)
|
79 |
+
delta = datetime.timedelta(seconds=sanitized_total_seconds)
|
80 |
+
total_int_seconds = int(delta.total_seconds())
|
81 |
+
|
82 |
+
hours = total_int_seconds // 3600
|
83 |
+
remainder_seconds_after_hours = total_int_seconds % 3600
|
84 |
+
minutes = remainder_seconds_after_hours // 60
|
85 |
+
seconds_part = remainder_seconds_after_hours % 60
|
86 |
+
milliseconds = delta.microseconds // 1000
|
87 |
+
|
88 |
+
return f"{hours:02d}:{minutes:02d}:{seconds_part:02d},{milliseconds:03d}"
|
89 |
+
|
90 |
+
def generate_srt_content(segment_timestamps: list) -> str:
|
91 |
+
"""Generates SRT formatted string from segment timestamps."""
|
92 |
+
srt_content = []
|
93 |
+
for i, ts in enumerate(segment_timestamps):
|
94 |
+
start_time = format_srt_time(ts['start'])
|
95 |
+
end_time = format_srt_time(ts['end'])
|
96 |
+
text = ts['segment']
|
97 |
+
srt_content.append(str(i + 1))
|
98 |
+
srt_content.append(f"{start_time} --> {end_time}")
|
99 |
+
srt_content.append(text)
|
100 |
+
srt_content.append("")
|
101 |
+
return "\n".join(srt_content)
|
102 |
+
|
103 |
@spaces.GPU
|
104 |
def get_transcripts_and_raw_times(audio_path, session_dir):
|
105 |
if not audio_path:
|
106 |
gr.Error("No audio file path provided for transcription.", duration=None)
|
107 |
+
# Return an update to hide the buttons
|
108 |
+
return [], [], None, gr.DownloadButton(label="Download Transcript (CSV)", visible=False), gr.DownloadButton(label="Download Transcript (SRT)", visible=False)
|
109 |
|
110 |
vis_data = [["N/A", "N/A", "Processing failed"]]
|
111 |
raw_times_data = [[0.0, 0.0]]
|
112 |
processed_audio_path = None
|
113 |
csv_file_path = None
|
114 |
+
srt_file_path = None
|
115 |
original_path_name = Path(audio_path).name
|
116 |
audio_name = Path(audio_path).stem
|
117 |
|
118 |
+
# Initialize button states
|
119 |
+
csv_button_update = gr.DownloadButton(label="Download Transcript (CSV)", visible=False)
|
120 |
+
srt_button_update = gr.DownloadButton(label="Download Transcript (SRT)", visible=False)
|
121 |
+
|
122 |
try:
|
123 |
try:
|
124 |
gr.Info(f"Loading audio: {original_path_name}", duration=2)
|
|
|
126 |
duration_sec = audio.duration_seconds
|
127 |
except Exception as load_e:
|
128 |
gr.Error(f"Failed to load audio file {original_path_name}: {load_e}", duration=None)
|
129 |
+
return [["Error", "Error", "Load failed"]], [[0.0, 0.0]], audio_path, csv_button_update, srt_button_update
|
|
|
130 |
|
131 |
resampled = False
|
132 |
mono = False
|
|
|
138 |
resampled = True
|
139 |
except Exception as resample_e:
|
140 |
gr.Error(f"Failed to resample audio: {resample_e}", duration=None)
|
141 |
+
return [["Error", "Error", "Resample failed"]], [[0.0, 0.0]], audio_path, csv_button_update, srt_button_update
|
|
|
142 |
|
143 |
if audio.channels == 2:
|
144 |
try:
|
|
|
146 |
mono = True
|
147 |
except Exception as mono_e:
|
148 |
gr.Error(f"Failed to convert audio to mono: {mono_e}", duration=None)
|
149 |
+
return [["Error", "Error", "Mono conversion failed"]], [[0.0, 0.0]], audio_path, csv_button_update, srt_button_update
|
|
|
150 |
elif audio.channels > 2:
|
151 |
gr.Error(f"Audio has {audio.channels} channels. Only mono (1) or stereo (2) supported.", duration=None)
|
152 |
+
return [["Error", "Error", f"{audio.channels}-channel audio not supported"]], [[0.0, 0.0]], audio_path, csv_button_update, srt_button_update
|
|
|
153 |
|
154 |
if resampled or mono:
|
155 |
try:
|
|
|
161 |
gr.Error(f"Failed to export processed audio: {export_e}", duration=None)
|
162 |
if processed_audio_path and os.path.exists(processed_audio_path):
|
163 |
os.remove(processed_audio_path)
|
164 |
+
return [["Error", "Error", "Export failed"]], [[0.0, 0.0]], audio_path, csv_button_update, srt_button_update
|
|
|
165 |
else:
|
166 |
transcribe_path = audio_path
|
167 |
info_path_name = original_path_name
|
|
|
191 |
|
192 |
if not output or not isinstance(output, list) or not output[0] or not hasattr(output[0], 'timestamp') or not output[0].timestamp or 'segment' not in output[0].timestamp:
|
193 |
gr.Error("Transcription failed or produced unexpected output format.", duration=None)
|
194 |
+
# Return an update to hide the buttons
|
195 |
+
return [["Error", "Error", "Transcription Format Issue"]], [[0.0, 0.0]], audio_path, csv_button_update, srt_button_update
|
196 |
|
197 |
segment_timestamps = output[0].timestamp['segment']
|
198 |
csv_headers = ["Start (s)", "End (s)", "Segment"]
|
199 |
vis_data = [[f"{ts['start']:.2f}", f"{ts['end']:.2f}", ts['segment']] for ts in segment_timestamps]
|
200 |
raw_times_data = [[ts['start'], ts['end']] for ts in segment_timestamps]
|
201 |
|
202 |
+
# CSV file generation
|
|
|
203 |
try:
|
204 |
csv_file_path = Path(session_dir, f"transcription_{audio_name}.csv")
|
205 |
writer = csv.writer(open(csv_file_path, 'w'))
|
206 |
writer.writerow(csv_headers)
|
207 |
writer.writerows(vis_data)
|
208 |
print(f"CSV transcript saved to temporary file: {csv_file_path}")
|
209 |
+
csv_button_update = gr.DownloadButton(value=csv_file_path, visible=True, label="Download Transcript (CSV)")
|
|
|
210 |
except Exception as csv_e:
|
211 |
gr.Error(f"Failed to create transcript CSV file: {csv_e}", duration=None)
|
212 |
print(f"Error writing CSV: {csv_e}")
|
213 |
+
|
214 |
+
if segment_timestamps:
|
215 |
+
try:
|
216 |
+
srt_content = generate_srt_content(segment_timestamps)
|
217 |
+
srt_file_path = Path(session_dir, f"transcription_{audio_name}.srt")
|
218 |
+
with open(srt_file_path, 'w', encoding='utf-8') as f:
|
219 |
+
f.write(srt_content)
|
220 |
+
print(f"SRT transcript saved to temporary file: {srt_file_path}")
|
221 |
+
srt_button_update = gr.DownloadButton(value=srt_file_path, visible=True, label="Download Transcript (SRT)")
|
222 |
+
except Exception as srt_e:
|
223 |
+
gr.Warning(f"Failed to create transcript SRT file: {srt_e}", duration=5)
|
224 |
+
print(f"Error writing SRT: {srt_e}")
|
225 |
|
226 |
gr.Info("Transcription complete.", duration=2)
|
227 |
+
return vis_data, raw_times_data, audio_path, csv_button_update, srt_button_update
|
|
|
228 |
|
229 |
except torch.cuda.OutOfMemoryError as e:
|
230 |
error_msg = 'CUDA out of memory. Please try a shorter audio or reduce GPU load.'
|
231 |
print(f"CUDA OutOfMemoryError: {e}")
|
232 |
gr.Error(error_msg, duration=None)
|
233 |
+
return [["OOM", "OOM", error_msg]], [[0.0, 0.0]], audio_path, csv_button_update, srt_button_update
|
|
|
234 |
|
235 |
except FileNotFoundError:
|
236 |
error_msg = f"Audio file for transcription not found: {Path(transcribe_path).name}."
|
237 |
print(f"Error: Transcribe audio file not found at path: {transcribe_path}")
|
238 |
gr.Error(error_msg, duration=None)
|
239 |
+
return [["Error", "Error", "File not found for transcription"]], [[0.0, 0.0]], audio_path, csv_button_update, srt_button_update
|
|
|
240 |
|
241 |
except Exception as e:
|
242 |
error_msg = f"Transcription failed: {e}"
|
|
|
244 |
gr.Error(error_msg, duration=None)
|
245 |
vis_data = [["Error", "Error", error_msg]]
|
246 |
raw_times_data = [[0.0, 0.0]]
|
247 |
+
return vis_data, raw_times_data, audio_path, csv_button_update, srt_button_update
|
|
|
248 |
finally:
|
249 |
# --- Model Cleanup ---
|
250 |
try:
|
|
|
382 |
gr.Markdown("<p><strong style='color: #FF0000; font-size: 1.2em;'>Transcription Results (Click row to play segment)</strong></p>")
|
383 |
|
384 |
# Define the DownloadButton *before* the DataFrame
|
385 |
+
with gr.Row():
|
386 |
+
download_btn_csv = gr.DownloadButton(label="Download Transcript (CSV)", visible=False)
|
387 |
+
download_btn_srt = gr.DownloadButton(label="Download Transcript (SRT)", visible=False)
|
388 |
|
389 |
vis_timestamps_df = gr.DataFrame(
|
390 |
headers=["Start (s)", "End (s)", "Segment"],
|
|
|
399 |
mic_transcribe_btn.click(
|
400 |
fn=get_transcripts_and_raw_times,
|
401 |
inputs=[mic_input, session_dir],
|
402 |
+
outputs=[vis_timestamps_df, raw_timestamps_list_state, current_audio_path_state, download_btn_csv, download_btn_srt],
|
403 |
api_name="transcribe_mic"
|
404 |
)
|
405 |
|
406 |
file_transcribe_btn.click(
|
407 |
fn=get_transcripts_and_raw_times,
|
408 |
inputs=[file_input, session_dir],
|
409 |
+
outputs=[vis_timestamps_df, raw_timestamps_list_state, current_audio_path_state, download_btn_csv, download_btn_srt],
|
410 |
api_name="transcribe_file"
|
411 |
)
|
412 |
|