import os
import openai
import gradio as gr
from youtube_transcript_api import YouTubeTranscriptApi
from urllib.parse import urlparse, parse_qs
from requests.structures import CaseInsensitiveDict
openai.api_key = os.getenv("OPENAI_KEY")
def Prompt_T(context, lang):
prompt = """I want you to act as a content writer who is working with youtube video transcript. Summarise the following text in 50 words:
=========
"""+ context +"""
=========
Answer:"""
if (lang=="ru"):prompt = """Я хочу, чтобы вы выступили в роли автора контента, который взят с транскрипт youtube видео, его нужно преобразовать в читаемый вид. Резюмируйте следующий текст в 40 слов:
=========
"""+ context +"""
=========
Ответ:"""
if (lang=="uk"):prompt = """Я хочу, щоб ви виступили в ролі автора контенту, який узятий з транскрипту youtube відео, його треба перетворити у читабельний вигляд. Резюмуйте наступний текст у 50 слів:
=========
"""+ context +"""
=========
Ответ:"""
return prompt
def convert_seconds(seconds):
seconds = round(seconds)
minutes = seconds // 60
hours = minutes // 60
minutes = minutes % 60
seconds = seconds % 60
if (int(hours) > 0 ): time_m= str(hours) + ":" + str(minutes) + ":" + str(seconds)
else : time_m = str(minutes) + ":" + str(seconds)
return time_m
def get_transcript(video_id, lang_video, chunk_size):
global final_string
transcript_list = YouTubeTranscriptApi.list_transcripts(video_id)
t = YouTubeTranscriptApi.get_transcript(video_id,languages=[lang_video])
finalString = ""
result_text_duration=[]
text_duration=""
prev_start=0.0
for item in t:
text = item['text']
finalString += text + " "
text_duration = text_duration + " " + item['text']
if (len(text_duration)>chunk_size):
new_item = {'text': prev_text_duration, 'start': prev_start}
result_text_duration.append(new_item)
text_duration=""
prev_start=item['start']
prev_text_duration=text_duration
new_item = {'text': text_duration, 'start': prev_start}
result_text_duration.append(new_item)
return result_text_duration
def split_string(string, chunk_size):
return [string[i:i+chunk_size] for i in range(0, len(string), chunk_size)]
def gpt_api (input_text):
completion = openai.ChatCompletion.create(
model="gpt-3.5-turbo",
messages=[ {"role": "system", "content": input_text} ]
)
response = completion.choices[0].message.content
return response
def generate_video_html(video_url, request: gr.Request):
my_v = ""
if (video_url =="") and (my_v == ""): video_url ="https://youtube.com/watch?v=PQBYZDyDBrY"
#похоже ли video_url на номальну ссылку
if "youtube.com/watch?v=" in video_url: x=111
else: return "Неверный URL"
#Пробуем извлеч video_id пока на английском
video_id = video_url[-11:]
html_embed='<iframe width="450" height="250" src="https://www.youtube.com/embed/'+ video_id +'" title="YouTube video player" frameborder="0" allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share" allowfullscreen name="video_yt" ></iframe>'
return html_embed
def generate(video_url, request: gr.Request):
my_v = ""
#Если две переменные пустые, то показываем базовую страницу с рекламой
if (video_url =="") and (my_v == ""):
html_embed='<div><br> An easy way to get video descriptions If you are on YouTube itself, simply add "zxc" in front of YouTube to the videos address.</div>'
summarize=""
return summarize
#похоже ли video_url на номальну ссылку
if "youtube.com/watch?v=" in video_url: x=111
else: return "Неверный URL"
#Пробуем извлеч video_id пока на английском
video_id = video_url[-11:]
print("video_id=", video_id)
try:
transcript_list = YouTubeTranscriptApi.list_transcripts(video_id)
except Exception as e:
return "No access for transcript "
lang_video="en"
for transcript in transcript_list:
if (transcript.is_generated == True): lang_video = transcript.language_code
print ("transcript.language_code=", transcript.language_code)
try:
t = YouTubeTranscriptApi.get_transcript(video_id,languages=[lang_video])
# do something with the transcript
except Exception as e:
return "No access for transcript"
finalString = ""
for item in t:
text = item['text']
finalString += text + " "
print ("===============================================")
#print("Transcript:",finalString)
print("Transcript lenght:",len(finalString))
print ("===============================================")
input_string = finalString
chunk_size = 4000
if (lang_video=="ru"): chunk_size = 2000
if (lang_video=="uk"): chunk_size = 2000
result_list = split_string(input_string, chunk_size)
result_list= text_video = get_transcript(video_id, lang_video, chunk_size)
final_answer_gpt=""
count= 0
img_wait='<img src="https://huggingface.co/spaces/LaoCzi/YouTube_Summarize2/resolve/main/22.gif">'
html_content="<br>"+img_wait
yield html_content
for item in result_list:
context = item['text']
time_text = str(convert_seconds(item['start']))
time_seconds = str(round(item['start']))
input_gpt = Prompt_T(context,lang_video)
time_url='<a href="//www.youtube.com/embed/'+ str(video_id) + '?rel=0&autoplay=1&start='+time_seconds +'" target="video_yt">'+ time_text+'</a>'
final_answer_gpt = final_answer_gpt +"<p>" + time_url +" " + gpt_api (input_gpt)+"</p>"
html_content="<h6>"+"<br>"+final_answer_gpt+img_wait
yield html_content
html_content="<h6>"+"<br>"+final_answer_gpt+"</h6>"
yield html_content
title = "YouTube Summorize (en,ua,ru)"
css="""
footer {visibility: hidden}
.gradio-container {padding-top: 100px}
"""
with gr.Blocks(css=css, title=title) as demo:
gr.HTML("<h3>A simple way to summarize YouTube video</h3>")
with gr.Row():
with gr.Column():
input_d = gr.Textbox(label="YouTube video URL", placeholder="https://www.youtube.com/watch?v=XXXXXXXX", value="")
greet_btn = gr.Button("Summarise")
dt_2 = gr.outputs.HTML()
dt_1 = gr.outputs.HTML()
dt =[dt_1, dt_2]
greet_btn.click(generate_video_html, inputs=input_d, outputs=dt_2)
greet_btn.click(generate, inputs=input_d, outputs=dt_1)
demo.load(generate_video_html, inputs=input_d, outputs=dt_2)
demo.load(generate, inputs=input_d, outputs=dt_1)
demo.queue()
demo.launch(debug=True, share=False)