Spaces:
Build error
Build error
Commit
·
6fa0583
1
Parent(s):
32c6ac4
Upload whisper_youtube.py
Browse files- whisper_youtube.py +75 -0
whisper_youtube.py
ADDED
|
@@ -0,0 +1,75 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# -*- coding: utf-8 -*-
|
| 2 |
+
"""whisper_youtube.ipynb
|
| 3 |
+
|
| 4 |
+
Automatically generated by Colaboratory.
|
| 5 |
+
|
| 6 |
+
Original file is located at
|
| 7 |
+
https://colab.research.google.com/drive/1spmA-7Un5TA6ahuCeO62BUS_ME6zPuUx
|
| 8 |
+
|
| 9 |
+
# Using gradio for making a nice UI.
|
| 10 |
+
Youtube link version.
|
| 11 |
+
|
| 12 |
+
Installing requirements.
|
| 13 |
+
"""
|
| 14 |
+
|
| 15 |
+
!pip install gradio
|
| 16 |
+
!pip install git+https://github.com/huggingface/transformers
|
| 17 |
+
|
| 18 |
+
!pip install pytube
|
| 19 |
+
|
| 20 |
+
from pytube import YouTube
|
| 21 |
+
from transformers import pipeline
|
| 22 |
+
import gradio as gr
|
| 23 |
+
import os
|
| 24 |
+
|
| 25 |
+
from transformers import WhisperProcessor
|
| 26 |
+
|
| 27 |
+
processor = WhisperProcessor.from_pretrained("openai/whisper-small", language="Galician", task="transcribe")
|
| 28 |
+
|
| 29 |
+
|
| 30 |
+
from transformers import WhisperTokenizer
|
| 31 |
+
|
| 32 |
+
tokenizer = WhisperTokenizer.from_pretrained("openai/whisper-small", language="Galician", task="transcribe")
|
| 33 |
+
|
| 34 |
+
"""## Building a Demo
|
| 35 |
+
|
| 36 |
+
Now that we've fine-tuned our model we can build a demo to show
|
| 37 |
+
off its ASR capabilities! We'll make use of 🤗 Transformers
|
| 38 |
+
`pipeline`, which will take care of the entire ASR pipeline,
|
| 39 |
+
right from pre-processing the audio inputs to decoding the
|
| 40 |
+
model predictions.
|
| 41 |
+
|
| 42 |
+
Running the example below will generate a Gradio demo where can input audio to
|
| 43 |
+
our fine-tuned Whisper model to transcribe the corresponding text:
|
| 44 |
+
"""
|
| 45 |
+
|
| 46 |
+
pipe = pipeline(model="Victorlopo21/whisper-medium-gl-30") # change to "your-username/the-name-you-picked"
|
| 47 |
+
|
| 48 |
+
def get_audio(url):
|
| 49 |
+
yt = YouTube(url)
|
| 50 |
+
video = yt.streams.filter(only_audio=True)[1]
|
| 51 |
+
out_file=video.download(output_path=".")
|
| 52 |
+
base, ext = os.path.splitext(out_file)
|
| 53 |
+
new_file = base+'.wav'
|
| 54 |
+
os.rename(out_file, new_file)
|
| 55 |
+
a = new_file
|
| 56 |
+
return a
|
| 57 |
+
|
| 58 |
+
def transcribe_url(url):
|
| 59 |
+
text = pipe(get_audio(url))['text']
|
| 60 |
+
return text
|
| 61 |
+
|
| 62 |
+
iface = gr.Interface(
|
| 63 |
+
fn=transcribe_url,
|
| 64 |
+
inputs='text',
|
| 65 |
+
outputs="text",
|
| 66 |
+
title="Whisper Medium Galician",
|
| 67 |
+
description="Realtime demo for Galician speech recognition using a fine-tuned Whisper medium model.",
|
| 68 |
+
)
|
| 69 |
+
|
| 70 |
+
iface.launch(debug=True)
|
| 71 |
+
|
| 72 |
+
# Short youtube video to hear
|
| 73 |
+
# https://www.youtube.com/watch?v=Z2SjeZJZi6s&ab_channel=rimc7
|
| 74 |
+
|
| 75 |
+
# TO TRY
|