Spaces:
Sleeping
Sleeping
File size: 2,914 Bytes
638be90 6e941de 638be90 3b83cb7 d2639f0 e129c20 d2639f0 ae41ebd d2639f0 ae41ebd 8078e71 e5be212 d2639f0 ae41ebd d2639f0 e5be212 e129c20 638be90 d834007 638be90 d834007 2fd9b5b 638be90 ae41ebd e129c20 d834007 e129c20 e0aae73 ae41ebd e129c20 638be90 ae41ebd 0c9ce61 638be90 ae41ebd 638be90 |
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 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 |
import gradio as gr
from pytube import YouTube
import random
import requests,json
def create(prompt,openai_key):
headers = {
'Content-Type': 'application/json',
'Authorization': f'Bearer {openai_key}',
}
data = {
"model": "text-davinci-003",
"prompt": prompt,
"temperature": 0.7,
"max_tokens": 1024,
"top_p": 1.0,
"frequency_penalty": 0.0,
"presence_penalty": 0.0
}
url = 'https://api.openai.com/v1/completions'
r = requests.post(url,headers=headers,
data=json.dumps(data))
return r.json()
def convert(res,openai_key):
headers = {
'accept': 'application/json',
'x-gladia-key': '89b0adf5-fb2c-48ba-8a66-76b02827fd14',
# requests won't add a boundary if this header is set when you pass files=
# 'Content-Type': 'multipart/form-data',
}
data = res.json()
prediction = data['prediction']
content = []
for x in prediction:
content.append(x['transcription'])
auido_txt = '\n'.join(content)
prompt = f"将下面的内容，总结10条要点出来，\n{auido_txt}"
open_ai_res = create(prompt,openai_key)
answer = open_ai_res['choices'][0]['text']
res_content = f'音频内容：\n{auido_txt}\nGPT3总结的要点：\n{answer}'
return res_content
def get_audio(url):
yt = YouTube(url)
audio_file = f'{random.randint(10000,100000)}.mp4'
print(f'{url} {audio_file} start get audio ...')
yt.streams.filter(only_audio=True)[0].download(filename=audio_file)
print('aodio over ..')
return audio_file
def get_transcript(url,openai_key):
audio_file = get_audio(url)
# audio_file = 'tmp.mp4'
files = {
'audio': (f"{audio_file}", open(f'{audio_file}', 'rb'), 'video/mp4'),
'language': (None, 'english'),
'language_behaviour': (None, 'automatic single language'),
}
response = requests.post('https://api.gladia.io/audio/text/audio-transcription/', headers=headers, files=files)
print(response.text)
return convert(response,openai_key)
with gr.Blocks() as demo:
with gr.Row():
with gr.Column():
with gr.Row():
url = gr.Textbox(placeholder='Youtube video URL', label='URL')
openai_key = gr.Textbox(placeholder='Your openai key', label='key')
with gr.Row():
gr.Markdown("Larger models are more accurate, but slower. For 1min video, it'll take ~30s (tiny), ~1min (base), ~3min (small), ~5min (medium), etc.")
transcribe_btn = gr.Button('Transcribe')
with gr.Column():
outputs = gr.Textbox(placeholder='Transcription of the video', label='Transcription')
transcribe_btn.click(get_transcript, inputs=[url,openai_key], outputs=outputs)
demo.launch(debug=True)
|