2023年12月2日发(作者：华为v30价格)

OpenAI ChatGPT实现AI代你听播客python代码

import os

import openai

from pydub import AudioSegment

from _models import ChatOpenAI

from _splitter import SpacyTextSplitter

from llama_index import GPTListIndex, LLMPredictor, ServiceContext,

SimpleDirectoryReader

from llama__parser import SimpleNodeParser

# 音频总结

def audio_summary():

podcast

_mp3("./data/3")

# PyDub handles time in milliseconds

ten_minutes = 5 * 60 * 1000

total_length = len(podcast)

start = 0

index = 0

= while start < total_length:

end = start + ten_minutes

if end < total_length:

chunk = podcast[start:end]

else:

chunk = podcast[start:]

with open(f"./data/podcast_clip_{index}.mp3", "wb") as f:

(f, format="mp3")

start = end

index += 1

prompt = "这是一段Onboard播客"

for i in range(index):

clip = f"./data/podcast_clip_{i}.mp3"

audio_file= open(clip, "rb")

transcript

audio_file,prompt=prompt)

# mkdir ./data/transcripts if not exists

if not ("./data/transcripts"):

rs("./data/transcripts")

# write to file

with open(f"./data/transcripts/podcast_clip_{i}.txt", "w") as f:

= ribe("whisper-1", (transcript['text'])

# get last sentence of the transcript

sentences = transcript['text'].split("。")

prompt = sentences[-1]

llm_predictor = LLMPredictor(llm=ChatOpenAI(temperature=0,

model_name="gpt-3.5-turbo", max_tokens=1024))

text_splitter

chunk_size=2048)

parser = SimpleNodeParser(text_splitter=text_splitter)

documents = SimpleDirectoryReader('./data/transcripts').load_data()

nodes = _nodes_from_documents(documents)

service_context = _defaults(llm_predictor=llm_predictor)

list_index = GPTListIndex(nodes=nodes, service_context=service_context)

response = list_("请你用中文总结一下我们的播客内容:",

response_mode="tree_summarize")

print(response)

_key = "YOUR API KEY"

audio_summary()

= SpacyTextSplitter(pipeline="zh_core_web_sm",