Spaces:
Running
Running
yoru_tomosu
commited on
Commit
•
346ee80
1
Parent(s):
d8f80e1
Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,47 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
!pip install -q git+https://github.com/openai/whisper.git
|
2 |
+
!pip install -q gradio
|
3 |
+
!pip install -q deepl
|
4 |
+
!pip install -q requests
|
5 |
+
|
6 |
+
import whisper
|
7 |
+
model = whisper.load_model("base")
|
8 |
+
|
9 |
+
import deepl
|
10 |
+
deepl_auth_key = os.environ["ElevenLabs_API"]
|
11 |
+
|
12 |
+
def translate(text, target_lang):
|
13 |
+
translator = deepl.Translator(deepl_auth_key)
|
14 |
+
translated_text = translator.translate_text(text, target_lang=target_lang)
|
15 |
+
return translated_text
|
16 |
+
|
17 |
+
def transcribe(audio):
|
18 |
+
|
19 |
+
# load audio and pad/trim it to fit 30 seconds
|
20 |
+
audio = whisper.load_audio(audio)
|
21 |
+
audio = whisper.pad_or_trim(audio)
|
22 |
+
|
23 |
+
# make log-Mel spectrogram and move to the same device as the model
|
24 |
+
mel = whisper.log_mel_spectrogram(audio).to(model.device)
|
25 |
+
|
26 |
+
# detect the spoken language
|
27 |
+
_, probs = model.detect_language(mel)
|
28 |
+
print(f"Detected language: {max(probs, key=probs.get)}")
|
29 |
+
detect_lang = max(probs, key=probs.get)
|
30 |
+
|
31 |
+
# decode the audio
|
32 |
+
options = whisper.DecodingOptions()
|
33 |
+
result = whisper.decode(model, mel, options)
|
34 |
+
|
35 |
+
|
36 |
+
translated_text = translate(result.text, "JA")
|
37 |
+
return translated_text
|
38 |
+
|
39 |
+
import gradio as gr
|
40 |
+
|
41 |
+
title = 'Video Translator'
|
42 |
+
|
43 |
+
inputs = gr.Video()
|
44 |
+
outputs = gr.Text()
|
45 |
+
interface = gr.Interface(title=title, fn=transcribe, inputs=inputs, outputs=outputs)
|
46 |
+
interface.launch(debug=True)
|
47 |
+
|