Build a YouTube AI Video Summarizer with LLMs, OpenAI, and Streamlit

The YouTube AI Video Summarizer project is architecturally divided into three primary, interconnected components, each responsible for a distinct phase of the summarization process. These components work in synergy to deliver a seamless user experience: 1. **`scrape_youtube.py`**: This script is dedicated to the initial data acquisition phase. Its primary function is to extract essential information from YouTube videos, including metadata such as the video title and channel name, the full video transcript, and the video's thumbnail image. 2. **`summarize_text.py`**: Once the raw transcript data is collected, this script takes over. It utilizes the OpenAI API to process the extracted transcript, generating a concise summary and a list of key takeaways. This component is crucial for distilling the core message of the video. 3. **`app.py`**: This script serves as the user interface layer. Built with Streamlit, it provides a simple and intuitive web application where users can input YouTube URLs, select their desired output language, and trigger the summarization process. It orchestrates the calls to the other two components and displays the results in a user-friendly format.

With the video transcript successfully extracted, the next critical phase is to condense this information into a manageable summary. This is achieved using the `summarize_text.py` script, which integrates with the OpenAI API. For security and best practices, the OpenAI API key should be configured as an environment variable before running this script. This is typically done via the command line: `export OPENAI_API_KEY='your-api-key-here'`. The `summarize_text(text, lang='en')` function is the core of this component. It initializes the OpenAI client using the API key retrieved from the environment. A carefully crafted prompt is then sent to the OpenAI API. This prompt instructs the model to provide a summary and a list of key takeaways in a specified language (`lang`), which defaults to English. The prompt is structured to ensure the output is clearly delineated into a 'Summary' section and a 'Key Takeaways' section, presented as a bulleted list. The function utilizes the `gpt-3.5-turbo` model for summarization. The article notes that while `gpt-4-turbo` offers superior performance, it comes with a trade-off of slower inference times. The response from the OpenAI API, containing the generated summary and key takeaways, is then parsed to extract the content, which is subsequently returned by the function. This process effectively transforms raw transcript data into valuable, summarized insights.

The YouTube AI Video Summarizer seamlessly integrates its three core components to provide a fluid user experience. The workflow begins when a user interacts with the Streamlit web interface, initiating the summarization process. 1. **User Input**: The user pastes a YouTube URL into the provided text input field and selects their preferred output language (English, Spanish, or Korean). 2. **Trigger Summarization**: Upon clicking the 'Summarize' button, the `app.py` script springs into action. 3. **Extract Video ID and Metadata**: The application first calls `extract_video_id` to parse the URL and obtain the unique video identifier. This ID is then used by `extract_metadata` to fetch the video's title and channel name, which are subsequently displayed to the user. 4. **Download and Display Thumbnail**: The `get_thumbnail_from_url` function is invoked, using the video ID to download the video's thumbnail. This image is then presented to the user within the application interface. 5. **Retrieve and Summarize Transcript**: The `get_transcript_from_url` function uses the video ID to fetch the video's transcript. This raw transcript text is then passed to the `summarize_transcript` function, which in turn calls the `summarize_text` function from the `summarize_text.py` script. This function communicates with the OpenAI API, using the provided transcript and the selected language to generate a summary and key takeaways. 6. **Display Results**: Finally, the generated summary, formatted with a 'Summary' section and 'Key Takeaways' as bullet points, is displayed prominently in the Streamlit application. The entire process is designed to be efficient, providing users with quick access to the essential information contained within YouTube videos.