The code I used in this video: https://youtu.be/lNdpu6u9ZYM

transcribe-and-summarize-yt-video-audio-for-kiecodes-gpt.py

import copy
import json
import os

from openai import OpenAI

client = OpenAI(
    api_key="<Your api key here>"
)

prompt = """
    Transcribe the following audio from videos of the youtube channel "Kie Codes" into English.
"""

metadata = [
    {
        "url": "https://youtu.be/KdelZvfcPfk?si=IZVJVsinerSj6tN8",
        "title": "5 Productivity Tips for Programmers",
    },
    {
        "url": "https://youtu.be/Hlp-9cdImSM?si=HO2B24MerwuS7B0b",
        "title": "How to commit better with Git",
    },
    {
        "url": "https://youtu.be/EvpZkdkp-v0?si=4gFG9XB7Th9I42kY",
        "title": "How to use GIT HOOKS for better COMMITS (PYTHON CODE INCLUDED)",
    },
    {
        "url": "https://youtu.be/T0qWZ7St_GE?si=yGyhL-9XpqVtrHr3",
        "title": "What Programming Language should I learn first 2022?"
    },
    {
        "url": "https://youtu.be/211tiIqZ-58?si=mvKAcpmqVd5juaFX",
        "title": "Learn faster to code in a new programming language"
    },
    {
        "url": "https://youtu.be/JF4z2u2ftv0?si=jsflpLbFDCry91nG",
        "title": "Learn programming efficiently (5 Tips)"
    },
    {
        "url": "https://youtu.be/zombLkjem00?si=Fe1-2vX0LzQGVNi_",
        "title": "5 Things I wish I knew before becoming a software engineer"
    },
    {
        "url": "https://youtu.be/uQj5UNhCPuo?si=VfJYXM7JPgP2zQG6",
        "title": "Genetic Algorithms Explained By Example"
    },
    {
        "url": "https://youtu.be/nhT56blfRpE?si=gd9jZQsxTjtnWWVt",
        "title": "Genetic Algorithm from Scratch in Python (tutorial with code)"
    },
    {
        "url": "https://youtu.be/aOsET8KapQQ?si=yC9o95_5iapNJxli",
        "title": "Genetic Algorithm in Python generates Music (code included)"
    },
    {
        "url": "https://youtu.be/5y7pQaP-5Qw?si=A3b7488gx30mUpQU",
        "title": "Why you should use Type Hints in Python - Are type hints worth it?"
    },
    {
        "url": "https://youtu.be/a1NLvZ5rgvQ?si=zJuviwEn9d0MsJVr",
        "title": "How to create a pre-commit git hook for your python type checker?",
    },
    {
        "url": "https://youtu.be/yScuF1UgGU0?si=kH0NoV-a69KTeKYx",
        "title": "How to use python type hinting?",
    },
    {
        "url": "https://youtu.be/51EoNgwoaTo?si=JgIqDZC2SbHKeHjr",
        "title": "Neural Network from Scratch in Python",
    },
    {
        "url": "https://youtu.be/P8Xrj70qtyo?si=D8dsXi-sPU9FNJ72",
        "title": "Neural Networks Matrix Math and NumPy"
    },
    {
        "url": "https://youtu.be/pdyyQ-w_x0I?si=BRk6Si0JJZCC1ste",
        "title": "When did you learn to program? (Ask Kie #1)",
    },
    {
        "url": "https://youtu.be/GAFh2Z5VtgM?si=Ff4fFfDDpnCoKVvq",
        "title": "Coding an NFT crypto collectible in 3 days (DAY 1)",
    },
    {
        "url": "https://youtu.be/75D0JjX7EZg?si=zRE4sPjnJ92OmuOJ",
        "title": "How to implement an ERC721 Token and connect it to OpenSea (DAY 2)",
    },
    {
        "url": "https://youtu.be/EnIrWNFwN-U?si=VNoOUYSQgWEbmo3K",
        "title": "Creating a dApp and migrating to the Ethereum Mainnet (DAY 3)"
    },
    {
        "url": "https://youtu.be/LW1i-axSoYE?si=N8g0HNtBw-U3PDb5",
        "title": "Random NFT pictures in under 100 lines of JavaScript"
    }
]

# Transcribe the audio
print("Transcribing audio")
result = ""
result_json = copy.deepcopy(metadata)
for filename in os.listdir("audio"):
    name = os.path.splitext(filename)[0]
    ext = os.path.splitext(filename)[1]
    if ext not in [".mp3", ".mp4"]:
        continue
    print("Transcribing: " + filename)
    if os.path.isfile("transcript/" + name + ".txt"):
        print("File already transcribed: " + name + ".txt -> Skipping")
    else:
        audio_file = open("audio/" + filename, "rb")
        transcript = client.audio.transcriptions.create(
            model="whisper-1",
            file=audio_file,
            response_format="text",
            prompt=prompt
        )
        with open(f"transcript/{name}.txt", "w") as f:
            f.write(transcript)
            f.flush()

    transcript = open("transcript/" + name + ".txt", "r").read()
    result += f"""
Video Title: {metadata[int(name)-1]["title"]}
Video URL: {metadata[int(name)-1]["url"]}
Transcript: {transcript}
---
    """
    result_json[int(name)-1]["transcript"] = transcript

    with open(f"final/video-{name}.json", "w") as f:
        json.dump(
            {
                "title": metadata[int(name)-1]["title"],
                "url": metadata[int(name)-1]["url"],
                "transcript": transcript
            },
            f,
            indent=2
        )
        f.flush()

# Put all transcripts into one text file with urls and titles
with open("kiecodes-transcripts.txt", "w") as f:
    f.write(result)
    f.flush()

# Put all transcripts into one json file with urls and titles
with open("kiecodes-videos.json", "w") as f:
    json.dump(result_json, f)
    f.flush()

# Put all transcripts into one json file with urls and titles (beautified)
with open("kiecodes-videos-beautified.json", "w") as f:
    json.dump(result_json, f, indent=2)
    f.flush()

# Summarize the transcript
print("Summarizing transcripts")
for filename in os.listdir("transcript"):
    name = os.path.splitext(filename)[0]
    print(f"Summarizing: {filename}")
    if os.path.isfile(f"summary/{filename}"):
        print(f"File already summarized: {filename} -> Skipping")
    else:
        with open(f"transcript/{filename}", "r") as f:
            transcript = f.read()

            response = client.chat.completions.create(
                model="gpt-3.5-turbo-16k",
                messages=[
                    {
                        "role": "system",
                        "content": """
                            You are a youtube script summarizer, which outputs a two sentence summary of a youtube video transcript.
                            The host in the video is Kie and is male. The channel name is Kie Codes.
                            The user will only provide you with a transcript of the full video.
                            You only output the two sentences that summary.
                        """
                    },
                    {
                        "role": "user",
                        "content": transcript
                    }
                ]
            )

        with open(f"summary/{filename}", "w") as f:
            f.write(response.choices[0].message.content)
            f.flush()

# Create a summary file with all titles, urls and summaries of all videos
result = ""
result_json = copy.deepcopy(metadata)
for summary_filename in os.listdir("summary"):
    name = os.path.splitext(summary_filename)[0]
    with open(f"summary/{summary_filename}", "r") as f:
        summary = f.read()
        result += f"""
Video Title: {metadata[int(name)-1]["title"]}
Video URL: {metadata[int(name)-1]["url"]}
Summary: {summary}

---
        """
        result_json[int(name)-1]["summary"] = summary
        result_json[int(name)-1]["transcript_file"] = f"video-{name}.json"

with open("kiecodes-summary.txt", "w") as f:
    f.write(result)
    f.flush()

with open("kiecodes-summary-beautified.json", "w") as f:
    json.dump(result_json, f, indent=2)
    f.flush()