Spaces:
Running
Running
Add an option to toggle timestamps
Browse filesIf Return Timestamps is true, the app transcribes audio/video files with timestamps in the SRT format.
app.py
CHANGED
@@ -22,13 +22,25 @@ pipe = pipeline(
|
|
22 |
device=device,
|
23 |
)
|
24 |
|
25 |
-
|
26 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
27 |
if inputs is None:
|
28 |
raise gr.Error("No audio file submitted! Please upload or record an audio file before submitting your request.")
|
29 |
|
30 |
-
text = pipe(inputs, batch_size=BATCH_SIZE, generate_kwargs={"task": task}, return_timestamps=
|
31 |
-
|
|
|
|
|
|
|
|
|
32 |
|
33 |
|
34 |
def _return_yt_html_embed(yt_url):
|
@@ -95,6 +107,7 @@ mf_transcribe = gr.Interface(
|
|
95 |
inputs=[
|
96 |
gr.inputs.Audio(source="microphone", type="filepath", optional=True),
|
97 |
gr.inputs.Radio(["transcribe", "translate"], label="Task", default="transcribe"),
|
|
|
98 |
],
|
99 |
outputs="text",
|
100 |
layout="horizontal",
|
@@ -113,13 +126,14 @@ file_transcribe = gr.Interface(
|
|
113 |
inputs=[
|
114 |
gr.inputs.Audio(source="upload", type="filepath", optional=True, label="Audio file"),
|
115 |
gr.inputs.Radio(["transcribe", "translate"], label="Task", default="transcribe"),
|
|
|
116 |
],
|
117 |
outputs="text",
|
118 |
layout="horizontal",
|
119 |
theme="huggingface",
|
120 |
-
title="Whisper Large V3: Transcribe Audio",
|
121 |
description=(
|
122 |
-
"Transcribe long-form microphone or audio inputs with the click of a button!
|
123 |
f" checkpoint [{MODEL_NAME}](https://huggingface.co/{MODEL_NAME}) and 🤗 Transformers to transcribe audio files"
|
124 |
" of arbitrary length."
|
125 |
),
|
@@ -130,7 +144,8 @@ yt_transcribe = gr.Interface(
|
|
130 |
fn=yt_transcribe,
|
131 |
inputs=[
|
132 |
gr.inputs.Textbox(lines=1, placeholder="Paste the URL to a YouTube video here", label="YouTube URL"),
|
133 |
-
gr.inputs.Radio(["transcribe", "translate"], label="Task", default="transcribe")
|
|
|
134 |
],
|
135 |
outputs=["html", "text"],
|
136 |
layout="horizontal",
|
|
|
22 |
device=device,
|
23 |
)
|
24 |
|
25 |
+
def chunks_to_srt(chunks):
|
26 |
+
srt_format = ""
|
27 |
+
for i, chunk in enumerate(chunks, 1):
|
28 |
+
start_time, end_time = chunk['timestamp']
|
29 |
+
start_time_hms = "{:02}:{:02}:{:02},{:03}".format(int(start_time // 3600), int((start_time % 3600) // 60), int(start_time % 60), int((start_time % 1) * 1000))
|
30 |
+
end_time_hms = "{:02}:{:02}:{:02},{:03}".format(int(end_time // 3600), int((end_time % 3600) // 60), int(end_time % 60), int((end_time % 1) * 1000))
|
31 |
+
srt_format += f"{i}\n{start_time_hms} --> {end_time_hms}\n{chunk['text']}\n\n"
|
32 |
+
return srt_format
|
33 |
+
|
34 |
+
def transcribe(inputs, task, return_timestamps):
|
35 |
if inputs is None:
|
36 |
raise gr.Error("No audio file submitted! Please upload or record an audio file before submitting your request.")
|
37 |
|
38 |
+
text = pipe(inputs, batch_size=BATCH_SIZE, generate_kwargs={"task": task}, return_timestamps=return_timestamps)
|
39 |
+
|
40 |
+
if return_timestamps:
|
41 |
+
return chunks_to_srt(result['chunks'])
|
42 |
+
else:
|
43 |
+
return result['text']
|
44 |
|
45 |
|
46 |
def _return_yt_html_embed(yt_url):
|
|
|
107 |
inputs=[
|
108 |
gr.inputs.Audio(source="microphone", type="filepath", optional=True),
|
109 |
gr.inputs.Radio(["transcribe", "translate"], label="Task", default="transcribe"),
|
110 |
+
gr.inputs.Checkbox(label="Return timestamps"),
|
111 |
],
|
112 |
outputs="text",
|
113 |
layout="horizontal",
|
|
|
126 |
inputs=[
|
127 |
gr.inputs.Audio(source="upload", type="filepath", optional=True, label="Audio file"),
|
128 |
gr.inputs.Radio(["transcribe", "translate"], label="Task", default="transcribe"),
|
129 |
+
gr.inputs.Checkbox(label="Return timestamps"),
|
130 |
],
|
131 |
outputs="text",
|
132 |
layout="horizontal",
|
133 |
theme="huggingface",
|
134 |
+
title="Whisper Large V3: Transcribe Audio/Video",
|
135 |
description=(
|
136 |
+
"Transcribe long-form microphone or audio inputs with the click of a button! The app uses the"
|
137 |
f" checkpoint [{MODEL_NAME}](https://huggingface.co/{MODEL_NAME}) and 🤗 Transformers to transcribe audio files"
|
138 |
" of arbitrary length."
|
139 |
),
|
|
|
144 |
fn=yt_transcribe,
|
145 |
inputs=[
|
146 |
gr.inputs.Textbox(lines=1, placeholder="Paste the URL to a YouTube video here", label="YouTube URL"),
|
147 |
+
gr.inputs.Radio(["transcribe", "translate"], label="Task", default="transcribe"),
|
148 |
+
gr.inputs.Checkbox(label="Return timestamps"),
|
149 |
],
|
150 |
outputs=["html", "text"],
|
151 |
layout="horizontal",
|