File size: 1,810 Bytes
b1377a3 cfdbf22 7c4667a d411a02 7c4667a 965e092 7c4667a 49ed7a0 7c4667a 49ed7a0 7c4667a cfdbf22 7c4667a d411a02 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 |
from pytube import YouTube
import os
import torch
from transformers import AutoModelForSpeechSeq2Seq, AutoProcessor, pipeline
import whisperx
from datasets import load_dataset
import os.path as osp
from mlxtend.file_io import find_files
from mlxtend.utils import Counter
import accelerate
import gc
import gradio as gr
# Definimos una función que se encarga de llevar a cabo las transcripciones
def URLToText(URL):
# url input from user
yt = YouTube(URL)
# extract only audio
video = yt.streams.filter(only_audio=True).first()
# check for destination to save file
destination = '.'
# download the file
out_file = video.download(output_path=destination)
# save the file
base, ext = os.path.splitext(out_file)
base = base.replace(" ", "")
new_file = base + '.mp3'
os.rename(out_file, new_file)
# Pasamos el auido a texto
device = "cuda:0" if torch.cuda.is_available() else "cpu"
torch_dtype = torch.float16 if torch.cuda.is_available() else torch.float32
model_id = "openai/whisper-medium"
model = AutoModelForSpeechSeq2Seq.from_pretrained(
model_id, torch_dtype=torch_dtype, low_cpu_mem_usage=True, use_safetensors=True
)
model.to(device)
processor = AutoProcessor.from_pretrained(model_id)
pipe = pipeline(
"automatic-speech-recognition",
model=model,
tokenizer=processor.tokenizer,
feature_extractor=processor.feature_extractor,
max_new_tokens=128,
chunk_length_s=30,
batch_size=16,
return_timestamps=True,
torch_dtype=torch_dtype,
device=device,
)
result = pipe(new_file)
return result["text"]
# Creamos la interfaz y la lanzamos.
gr.Interface(fn=URLToText, inputs=gr.inputs.Textbox(label="Video URL"), outputs=gr.outputs.Textbox(label="Transcripción")).launch(share=False) |