app.py

import gradio as gr
import os,secrets,requests
from basic_pitch.inference import predict_and_save
from urllib.parse import quote
os.system('chmod +x ./pget')





def downloadaud(query):
 filename=secrets.token_hex()
 # Step 1: Search for videos with the given query
 search_url = f"https://draw-8fj.begin.app/api/search/{quote(query)}"
 search_response = requests.get(search_url).json()
 print('1=>', search_response)
 # Step 2: Find the first video with duration less than 10 minutes and extract its ID
 video_id = None
 #search_response = sorted(search_response, key=lambda x: x["views"], reverse=True)
 for item in search_response:
 duration = item.get("duration_raw")
 if duration and len(duration.split(':'))< 3 and int(duration.split(':')[0])<10:
 video_id = item.get("id", {}).get("videoId")
 break
 print('1-r',video_id)
 # If no video with duration less than 10 minutes was found, return None
 if not video_id:
 return None

 # Step 3: Get the formats for the video and find the URL for the best audio-only format
 formats_url = f"https://draw-8fj.begin.app/api/info/{video_id}"
 formats = requests.get(formats_url)
 if formats.ok!=True:
 formats_url = f"https://draw-8fj-staging.begin.app/api/info/{video_id}"
 formats = requests.get(formats_url) 
 formats_response=formats.json()
 print(formats_response["formats"])
 best_audio_format = None
 for fmt in formats_response.get("formats", []):
 if fmt.get("hasVideo") is False and fmt.get("hasAudio") is True and fmt.get("container") == "mp4":
 if not best_audio_format or fmt.get("audioBitrate") > best_audio_format.get("audioBitrate"):
 best_audio_format = fmt
 print(best_audio_format)
 # If no suitable audio format was found, return None
 if not best_audio_format:
 return None

 aurl=best_audio_format["url"]
 print(aurl)
 command = f'./pget -o {filename}.mp4 -p 4 "{aurl}" '
 os.system(command)
 os.system(f'ffmpeg -i {filename}.mp4 -q:a 0 -map a {filename}.mp3')
 os.system(f'mv {filename}.mp3 /tmp/{filename}.mp3')
 res= predict(f'/tmp/{filename}.mp3')
 return res


def predict(ina):
 fn=ina
 predict_and_save(
 [ina],
 '/'.join(ina.split('/')[:-1]),
 True,
 True,
 False,
 False,
)
 ina='.'.join(ina.split('.')[:-1])
 
 ina=ina+'_basic_pitch'
 os.system(f'ffmpeg -i "{ina}.wav" -f mp3 -acodec libmp3lame -y "{ina}.mp3"')
 
 return f'{fn}',f'{ina}.mp3',f'{ina}.mid'
aud=gr.Textbox(label="yt query")
oar=gr.Audio(label="下载的音频")
oa=gr.Audio(label="预览midi")
pf=gr.File(label="下载midi")
demo = gr.Interface(fn=downloadaud, inputs=[aud], outputs=[oar,oa,pf])
demo.launch()