Spaces:
Running
on
T4
Running
on
T4
update test
Browse files
app.py
CHANGED
|
@@ -65,24 +65,19 @@ def _return_yt_html_embed(yt_url):
|
|
| 65 |
return HTML_str
|
| 66 |
|
| 67 |
|
| 68 |
-
|
| 69 |
-
|
| 70 |
-
|
| 71 |
-
'outtmpl': 'audio.mp3',
|
| 72 |
-
'noplaylist': True
|
| 73 |
-
}
|
| 74 |
|
| 75 |
-
|
| 76 |
-
|
| 77 |
-
|
| 78 |
-
except Exception as e:
|
| 79 |
-
return f"Error downloading audio: {str(e)}"
|
| 80 |
|
| 81 |
-
|
| 82 |
-
|
| 83 |
|
| 84 |
-
text =
|
| 85 |
-
return
|
| 86 |
|
| 87 |
|
| 88 |
demo = gr.Blocks()
|
|
@@ -90,8 +85,8 @@ demo = gr.Blocks()
|
|
| 90 |
mf_transcribe = gr.Interface(
|
| 91 |
fn=transcribe,
|
| 92 |
inputs=[
|
| 93 |
-
gr.
|
| 94 |
-
gr.
|
| 95 |
],
|
| 96 |
outputs="text",
|
| 97 |
title="NB-Whisper Demo",
|
|
@@ -106,20 +101,20 @@ mf_transcribe = gr.Interface(
|
|
| 106 |
yt_transcribe = gr.Interface(
|
| 107 |
fn=yt_transcribe,
|
| 108 |
inputs=[
|
| 109 |
-
gr.
|
| 110 |
-
gr.
|
| 111 |
],
|
| 112 |
-
examples=[["https://www.youtube.com/watch?v=mukeSSa5GKo"]],
|
| 113 |
outputs=["html", "text"],
|
| 114 |
title="Whisper Demo: Transcribe YouTube",
|
| 115 |
description=(
|
| 116 |
"Transcribe long-form YouTube videos with the click of a button! Demo uses the the fine-tuned checkpoint:"
|
| 117 |
-
f" [{MODEL_NAME}](https://huggingface.co/{MODEL_NAME}) and 🤗 Transformers to transcribe
|
| 118 |
" arbitrary length."
|
| 119 |
),
|
| 120 |
allow_flagging="never",
|
| 121 |
)
|
| 122 |
|
|
|
|
| 123 |
with demo:
|
| 124 |
gr.TabbedInterface([
|
| 125 |
mf_transcribe,
|
|
|
|
| 65 |
return HTML_str
|
| 66 |
|
| 67 |
|
| 68 |
+
@spaces.GPU
|
| 69 |
+
def yt_transcribe(yt_url, task):
|
| 70 |
+
html_embed_str = _return_yt_html_embed(yt_url)
|
|
|
|
|
|
|
|
|
|
| 71 |
|
| 72 |
+
with tempfile.TemporaryDirectory() as tmpdirname:
|
| 73 |
+
filepath = os.path.join(tmpdirname, "audio.mp3")
|
| 74 |
+
download_yt_audio(yt_url, filepath)
|
|
|
|
|
|
|
| 75 |
|
| 76 |
+
inputs = ffmpeg_read(filepath, pipe.feature_extractor.sampling_rate)
|
| 77 |
+
inputs = {"array": inputs, "sampling_rate": pipe.feature_extractor.sampling_rate}
|
| 78 |
|
| 79 |
+
text = pipe(inputs, batch_size=BATCH_SIZE, generate_kwargs={"task": task}, return_timestamps=True)["text"]
|
| 80 |
+
return html_embed_str, text
|
| 81 |
|
| 82 |
|
| 83 |
demo = gr.Blocks()
|
|
|
|
| 85 |
mf_transcribe = gr.Interface(
|
| 86 |
fn=transcribe,
|
| 87 |
inputs=[
|
| 88 |
+
gr.Audio(sources="microphone", type="filepath"),
|
| 89 |
+
gr.Radio(["transcribe", "translate"], label="Task", value="transcribe"),
|
| 90 |
],
|
| 91 |
outputs="text",
|
| 92 |
title="NB-Whisper Demo",
|
|
|
|
| 101 |
yt_transcribe = gr.Interface(
|
| 102 |
fn=yt_transcribe,
|
| 103 |
inputs=[
|
| 104 |
+
gr.Textbox(lines=1, placeholder="Paste the URL to a YouTube video here", label="YouTube URL"),
|
| 105 |
+
gr.Radio(["transcribe", "translate"], label="Task", value="transcribe"),
|
| 106 |
],
|
|
|
|
| 107 |
outputs=["html", "text"],
|
| 108 |
title="Whisper Demo: Transcribe YouTube",
|
| 109 |
description=(
|
| 110 |
"Transcribe long-form YouTube videos with the click of a button! Demo uses the the fine-tuned checkpoint:"
|
| 111 |
+
f" [{MODEL_NAME}](https://huggingface.co/{MODEL_NAME}) and 🤗 Transformers to transcribe video files of"
|
| 112 |
" arbitrary length."
|
| 113 |
),
|
| 114 |
allow_flagging="never",
|
| 115 |
)
|
| 116 |
|
| 117 |
+
|
| 118 |
with demo:
|
| 119 |
gr.TabbedInterface([
|
| 120 |
mf_transcribe,
|