Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
@@ -1,23 +1,27 @@
|
|
1 |
import gradio as gr
|
2 |
import whisper
|
3 |
-
import subprocess
|
4 |
-
import os
|
5 |
from pytube import YouTube
|
6 |
from fastapi import FastAPI, Response, Request
|
7 |
import yt_dlp
|
|
|
|
|
|
|
|
|
|
|
8 |
|
9 |
langs = ["None"] + sorted(list(whisper.tokenizer.LANGUAGES.values()))
|
10 |
model_size = list(whisper._MODELS.keys())
|
11 |
|
12 |
|
13 |
-
|
14 |
-
async def get_subtitle(url: str):
|
15 |
# Download the subtitle with download_subtitle()
|
16 |
-
subtitle_url = download_subtitle(url)
|
17 |
# Stream the subtitle as a response
|
18 |
-
return StreamingResponse(requests.get(subtitle_url, stream=True).iter_content(chunk_size=1024))
|
19 |
|
20 |
-
|
|
|
21 |
# Download subtitles if available
|
22 |
ydl_opts = {
|
23 |
'writesubtitles': True,
|
@@ -98,29 +102,33 @@ def format_timestamp(t):
|
|
98 |
mi = (t - int(t))*1000
|
99 |
return f"{int(hh):02d}:{int(mm):02d}:{int(ss):02d},{int(mi):03d}"
|
100 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
101 |
|
102 |
-
|
103 |
-
|
104 |
-
with gr.Row():
|
105 |
-
|
106 |
-
with gr.Column():
|
107 |
-
|
108 |
-
with gr.Row():
|
109 |
-
url = gr.Textbox(placeholder='Youtube video URL', label='URL')
|
110 |
-
|
111 |
-
with gr.Row():
|
112 |
-
|
113 |
-
model_size = gr.Dropdown(choices=model_size, value='tiny', label="Model")
|
114 |
-
lang = gr.Dropdown(choices=langs, value="None", label="Language (Optional)")
|
115 |
-
format = gr.Dropdown(choices=["None", ".srt"], value="None", label="Timestamps? (Optional)")
|
116 |
-
|
117 |
-
with gr.Row():
|
118 |
-
gr.Markdown("Larger models are more accurate, but slower. For 1min video, it'll take ~30s (tiny), ~1min (base), ~3min (small), ~5min (medium), etc.")
|
119 |
-
transcribe_btn = gr.Button('Transcribe')
|
120 |
-
|
121 |
-
with gr.Column():
|
122 |
-
outputs = gr.Textbox(placeholder='Transcription of the video', label='Transcription')
|
123 |
-
|
124 |
-
transcribe_btn.click(get_transcript, inputs=[url, model_size, lang, format], outputs=outputs)
|
125 |
|
126 |
-
|
|
|
|
1 |
import gradio as gr
|
2 |
import whisper
|
|
|
|
|
3 |
from pytube import YouTube
|
4 |
from fastapi import FastAPI, Response, Request
|
5 |
import yt_dlp
|
6 |
+
import uvicorn
|
7 |
+
|
8 |
+
CUSTOM_PATH = "/"
|
9 |
+
|
10 |
+
app = FastAPI()
|
11 |
|
12 |
langs = ["None"] + sorted(list(whisper.tokenizer.LANGUAGES.values()))
|
13 |
model_size = list(whisper._MODELS.keys())
|
14 |
|
15 |
|
16 |
+
|
17 |
+
#async def get_subtitle(url: str):
|
18 |
# Download the subtitle with download_subtitle()
|
19 |
+
#subtitle_url = download_subtitle(url)
|
20 |
# Stream the subtitle as a response
|
21 |
+
#return StreamingResponse(requests.get(subtitle_url, stream=True).iter_content(chunk_size=1024))
|
22 |
|
23 |
+
@app.get("/subtitle")
|
24 |
+
async def get_subtitle(url, lang='en'):
|
25 |
# Download subtitles if available
|
26 |
ydl_opts = {
|
27 |
'writesubtitles': True,
|
|
|
102 |
mi = (t - int(t))*1000
|
103 |
return f"{int(hh):02d}:{int(mm):02d}:{int(ss):02d},{int(mi):03d}"
|
104 |
|
105 |
+
def gradio_interface():
|
106 |
+
with gr.Blocks() as demo:
|
107 |
+
|
108 |
+
with gr.Row():
|
109 |
+
|
110 |
+
with gr.Column():
|
111 |
+
|
112 |
+
with gr.Row():
|
113 |
+
url = gr.Textbox(placeholder='Youtube video URL', label='URL')
|
114 |
+
|
115 |
+
with gr.Row():
|
116 |
+
|
117 |
+
model_size = gr.Dropdown(choices=model_size, value='tiny', label="Model")
|
118 |
+
lang = gr.Dropdown(choices=langs, value="None", label="Language (Optional)")
|
119 |
+
format = gr.Dropdown(choices=["None", ".srt"], value="None", label="Timestamps? (Optional)")
|
120 |
+
|
121 |
+
with gr.Row():
|
122 |
+
gr.Markdown("Larger models are more accurate, but slower. For 1min video, it'll take ~30s (tiny), ~1min (base), ~3min (small), ~5min (medium), etc.")
|
123 |
+
transcribe_btn = gr.Button('Transcribe')
|
124 |
+
|
125 |
+
with gr.Column():
|
126 |
+
outputs = gr.Textbox(placeholder='Transcription of the video', label='Transcription')
|
127 |
+
|
128 |
+
transcribe_btn.click(get_transcript, inputs=[url, model_size, lang, format], outputs=outputs)
|
129 |
|
130 |
+
#demo.launch(debug=True)
|
131 |
+
io = gr.Interface(gradio_interface)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
132 |
|
133 |
+
app = gr.mount_gradio_app(app, io, path=CUSTOM_PATH)
|
134 |
+
uvicorn.run(app, host="0.0.0.0", port=7860)
|