Document

3. Video to Annotated "how to" article

I want to record a video on a "how to" topic and turn it into an article, with text and images. For example, I will record a video about "how to make software mockups with GIMP" and then want it turned into a video. I want this to be a very media-rich article with lots and lots of screenshots, so the idea is that in the video I will explain what I am doing (eg "the first thing we do is open GIMP and click on new image and choose an image size of 1000x1000"), and then we will want to extract the transcription from the video. Then we review this transcription/timestamps and decide on the core steps of the how to task (eg step 1 - download GIMP.. step 2 - create a new image), and then we look at the timestamps mentioned and extract images at those specific timestamps (so if I say "let me open the image now" at 23.39 then we would extract an image frame from the video at that point). Then we will use my transcription file to set specific timings.

final-output-and-component-outputs

Thank you for the detailed explanation of your project. Let me address your questions one by one to ensure I have a clear understanding of the final output and the necessary components.

a) Final Output:
Based on your description, the final output would be a comprehensive, media-rich article in a web-friendly format (e.g., HTML, Markdown, or a CMS-specific format) that includes the following elements:
- The text content of the article, derived from the transcription of your "how to" video.
- Relevant images extracted from the video at specific timestamps corresponding to the key steps mentioned in the transcription.
- Proper formatting, such as headings, paragraphs, and image placement, to ensure a logical flow and easy readability.

b) Core Components/Outputs:
To generate the final output, the following components would be necessary:

1. Video Transcription:
   - A text file containing the full transcription of your "how to" video.
   - The transcription should include timestamps for each sentence or phrase to facilitate image extraction and synchronization.

2. Key Steps Identification:
   - A list of the core steps involved in the "how to" process, derived from analyzing the video transcription.
   - Each step should have a corresponding timestamp or range of timestamps.

3. Image Extraction:
   - A set of images extracted from the video at the specific timestamps mentioned in the transcription.
   - These images should be relevant to the key steps and visually illustrate the actions being described.
   - The images should be saved in a web-friendly format (e.g., JPEG, PNG) with appropriate file names.

4. Article Structure:
   - A structured outline or template for the article, organizing the content into sections, headings, and paragraphs.
   - The structure should follow a logical flow based on the identified key steps.

5. Integration and Formatting:
   - A script or program that combines the transcription text, extracted images, and article structure into a cohesive and well-formatted article.
   - The script should handle the placement of images alongside the relevant text based on the timestamps.
   - The output should be in a format suitable for web publishing (e.g., HTML, Markdown, or a CMS-specific format).

6. Timestamp Synchronization:
   - A mechanism to synchronize the text and images based on the timestamps from the transcription.
   - This could involve creating a data structure or file (e.g., JSON) that maps the timestamps to the corresponding text and image elements.

By breaking down the task into these specific components and outputs, an AI agent can systematically process the video, extract the necessary information, and generate a comprehensive, media-rich article that aligns with your vision.

claude 3 - extract final output

extract component outputs

b) Core Components/Outputs:
To generate the final output, the following components would be necessary:

Claude - research-ideas-and-input-variables

Thank you for the detailed explanation of your task. To ensure that the AI Agent can successfully complete the project, it's essential to clarify the final output and the necessary inputs. Here are some questions and considerations:

a) Questions to clarify the final output:

1. What is the desired format of the final article? Will it be a blog post, a step-by-step guide, or another specific format?
2. How long should the article be in terms of word count?
3. Should the article include a specific number of images, or should it have as many images as necessary to illustrate each step clearly?
4. Will the article require any additional elements, such as a table of contents, a summary, or a conclusion section?
5. Should the article be optimized for search engines (SEO) with specific keywords, meta descriptions, and alt tags for images?
6. Is there a specific platform where the article will be published (e.g., WordPress, Medium, or a custom website)?

b) Inputs the task should accept before launching the task completion process:

1. The video file containing the "how to" tutorial.
2. The desired title of the article.
3. A list of any specific keywords or phrases to include in the article for SEO purposes (if applicable).
4. The desired word count range for the article (e.g., 1000-1500 words).
5. The desired number of images to include in the article (if there's a specific requirement).
6. Any additional instructions or preferences for the article's structure, tone, or style.

By obtaining clear answers to these questions and ensuring that the necessary inputs are provided, the AI Agent will have a solid foundation to build out the automated workflow. This will help ensure that the final output meets your expectations and that the project is completed successfully.

GPT - research-ideas-and-input-variables

a) It's vital to accurately understand your end goal. Here are some clarifying questions:

1. What is the length of the final article you're envisioning? This could be in terms of word count, number of steps involved or number of sections. 
2. Do you have a specific quality or resolution you want for the screenshots? 
3. How detailed do you expect the steps in the article to be? Should they be descriptive paragraphs, or short and quick bullet points? 
4. Do you have any preferred structure for the final article? For instance, introduction, followed by the series of steps, then conclusion, etc.
5. Is there a specific style or tone you want for the article? Formal, conversational, humorous, etc. 
6. Would you like any specific elements included, such as an overview, a list of necessary tools to complete the task, difficulty level or time required?
7. What format should the final output be in? For example, should it be a Word document, or a blog post in HTML format, PDF, etc.?
8. If the task is recording a series of videos on different topics, how many articles do you want to generate from each video?

b) Here are the inputs that would be helpful to get started with the task:

1. The video file(s) that you have recorded and would like to convert into an article.
2. Any additional specifications or instructions that weren't covered in the video and you would like to include in the article.
3. If there is a specific timeframe within the video that the AI should pay attention to when getting screenshot(s) (although the AI should be capable of picking the frames when significant actions are being performed)
4. Any accompanying notes, scripts or outlines you used when recording the video.
5. The style guide or any specific language usage preferences. 
6. Your preference for the frequency and position of images within the article (whether they should be embedded within relevant portions of texts or set apart).

template-ideas

Thank you for the detailed explanation of your project and the importance of templates in defining a successful output. Based on your description, I suggest the following templates that could serve as useful references for your AI Agent:

1. Successful "how-to" articles from popular websites:
   - Visit websites like WikiHow, Instructables, or HowToGeek and search for well-received articles on topics similar to your "how to make software mockups with GIMP" video.
   - Analyze the structure, formatting, and content of these articles to understand what makes them effective and engaging.
   - Pay attention to the use of images, step-by-step instructions, and the overall flow of the article.

2. Documentation and tutorials from the official GIMP website:
   - Explore the official GIMP documentation and tutorials to see how they present information about using the software.
   - Take note of the language used, the level of detail provided, and the use of screenshots or illustrations to support the text.

3. Popular YouTube tutorials on using GIMP for mockups:
   - Search for highly-viewed and positively-rated YouTube tutorials on using GIMP for creating software mockups.
   - Analyze the video structure, pacing, and the balance between verbal explanations and visual demonstrations.
   - Examine the video descriptions and comments to identify what viewers find most helpful or engaging about these tutorials.

4. Successful blog posts or articles on creating software mockups:
   - Search for well-received blog posts or articles that cover the topic of creating software mockups, either with GIMP or other similar tools.
   - Study the use of images, code snippets (if applicable), and the overall presentation of the content.
   - Consider how these articles break down the process into understandable steps and provide useful tips or best practices.

5. README files or documentation from popular GitHub repositories related to software mockups:
   - Look for GitHub repositories that contain projects or resources related to creating software mockups.
   - Review the README files and any accompanying documentation to see how they describe the process and provide instructions to users.
   - Analyze the use of markdown formatting, code examples, and screenshots in these files.

By collecting and analyzing these templates, your AI Agent can gain valuable insights into what constitutes a successful output for your specific task. The Agent can then use this information to guide the creation of your media-rich article, ensuring that it incorporates the best practices and elements observed in these templates.

claude 3 - searches

Here are 5 specific keyphrases for you to search, to help refine the definition of what a successful output might look like for your task of creating a media-rich article based on a "how to make software mockups with GIMP" video:

1. "how to create a website mockup in gimp" wikihow
2. gimp official documentation "create mockup"
3. "gimp for ui design" tutorial youtube
4. "creating a software mockup in gimp" blog post
5. github readme "gimp mockup tutorial"

These keyphrases are designed to help you find:

1. A well-structured and detailed WikiHow article on creating website mockups using GIMP.
2. Official GIMP documentation or tutorials specifically related to creating mockups.
3. A popular YouTube tutorial video demonstrating how to use GIMP for UI design.
4. A comprehensive blog post that covers the process of creating software mockups in GIMP.
5. A GitHub repository with a README file that provides a step-by-step tutorial on creating mockups using GIMP.

By searching for these specific keyphrases and reviewing the top results, you should be able to find several high-quality templates that can help you refine your understanding of what makes a successful media-rich article on the topic of creating software mockups with GIMP. These templates will provide insights into the structure, content, and presentation of the information, which your AI Agent can then use to guide the creation of your own article.

subagent-summary

A) FULL TASK SUMMARY: Create a media-rich article from a "how to" video by extracting text and images based on the video transcription and timestamps.

B) FINAL TASK OUTPUT: An HTML file containing the formatted article with text content, headings, paragraphs, and relevant images extracted from the video. The images should be saved as separate files (JPEG or PNG) and referenced within the HTML file.

C) COMPONENTS OVERVIEW:
1. Video transcription with timestamps
2. Key steps identification
3. Image extraction at specific timestamps
4. Article structure and outline
5. Integration of text, images, and formatting
6. Timestamp synchronization

D) OVERVIEW OF SUBAGENTS:
1. Transcription Agent: Transcribes the video and generates a text file with timestamps.
2. Key Steps Agent: Analyzes the transcription and identifies the key steps and their corresponding timestamps.
3. Image Extraction Agent: Extracts relevant images from the video at the specified timestamps.
4. Article Structure Agent: Creates a structured outline for the article based on the key steps.
5. Integration Agent: Combines the text, images, and formatting into a cohesive HTML article.

E) SUBAGENT 1 - Transcription Agent:
- Name: VideoTranscriber
- Final Output: A text file containing the video transcription with timestamps for each sentence or phrase.
- Process:
  1. Use a speech recognition library (e.g., Google Speech-to-Text API or DeepSpeech) to transcribe the audio from the video.
  2. Split the transcription into sentences or phrases.
  3. Use a video processing library (e.g., OpenCV or FFmpeg) to extract the timestamps for each sentence or phrase.
  4. Save the transcription with timestamps as a plain text file.

F) SUBAGENT 2 - Key Steps Agent:
- Name: KeyStepsExtractor
- Final Output: A list of the core steps involved in the "how to" process, along with their corresponding timestamp ranges.
- Process:
  1. Analyze the transcription to identify key phrases or sentences that indicate a new step (e.g., "first," "next," "then," "finally").
  2. Extract the identified steps and their corresponding timestamp ranges.
  3. Save the list of key steps and timestamps as a JSON file.

G) SUBAGENT 3 - Image Extraction Agent:
- Name: ImageExtractor
- Final Output: A set of images extracted from the video at the specified timestamps, saved as JPEG or PNG files.
- Process:
  1. Use the list of key steps and timestamps from the KeyStepsExtractor.
  2. For each timestamp or timestamp range, use a video processing library (e.g., OpenCV or FFmpeg) to extract the corresponding frame(s) from the video.
  3. Save the extracted images as JPEG or PNG files with appropriate file names.

H) SUBAGENT 4 - Article Structure Agent:
- Name: ArticleStructureCreator
- Final Output: A structured outline for the article in JSON format.
- Process:
  1. Use the list of key steps from the KeyStepsExtractor.
  2. Create a hierarchical structure for the article, with headings, subheadings, and placeholders for text and images based on the key steps.
  3. Save the article structure as a JSON file.

I) SUBAGENT 5 - Integration Agent:
- Name: ArticleIntegrator
- Final Output: An HTML file containing the formatted article with text, images, and proper layout.
- Process:
  1. Load the transcription text, extracted images, and article structure.
  2. Combine the text and images according to the article structure, inserting the images at the appropriate locations based on the timestamps.
  3. Apply formatting and styling to the article using HTML tags and CSS.
  4. Save the final article as an HTML file, along with the extracted images in a separate folder.

subagent-summary-amended-for-actual-tasks

Thank you for providing the list of available tasks. Based on the updated information, here's the revised list of subagents and their corresponding tasks:

A) FULL TASK SUMMARY: Create a media-rich article from a "how to" video by extracting text and images based on the video transcription and timestamps.

E) SUBAGENT 1 - Transcription Agent:
- Name: VideoTranscriber
- Final Output: A text file containing the video transcription with timestamps for each sentence or phrase.
- Tasks:
  1. Download youtube video to mp4 (Task 43)
  2. Convert mp4 to mp3 (Task 44)
  3. Get text transcription of mp3 (SRT subtitles) (Task 32)
  4. Save content to server (Task 9) - Save the transcription as a text file on the server

F) SUBAGENT 2 - Key Steps Agent:
- Name: KeyStepsExtractor
- Final Output: A list of the core steps involved in the "how to" process, along with their corresponding timestamp ranges.
- Tasks:
  1. Access text files already saved (Task 10) - Access the transcription file
  2. GPT4 (Task 1) - Analyze the transcription to identify key phrases or sentences that indicate a new step
  3. GPT4 (Task 1) - Extract the identified steps and their corresponding timestamp ranges
  4. Save content to server (Task 9) - Save the list of key steps and timestamps as a JSON file

G) SUBAGENT 3 - Image Extraction Agent:
- Name: ImageExtractor
- Final Output: A set of images extracted from the video at the specified timestamps, saved as JPEG or PNG files.
- Tasks:
  1. Access text files already saved (Task 11) - Access the JSON file with key steps and timestamps
  2. Split/extract video by interval (Task 45) - For each timestamp or timestamp range, extract the corresponding frame(s) from the video
  3. Save content to server (Task 9) - Save the extracted images as JPEG or PNG files with appropriate file names

H) SUBAGENT 4 - Article Structure Agent:
- Name: ArticleStructureCreator
- Final Output: A structured outline for the article in JSON format.
- Tasks:
  1. Access text files already saved (Task 11) - Access the JSON file with key steps
  2. GPT4 (Task 1) - Create a hierarchical structure for the article, with headings, subheadings, and placeholders for text and images based on the key steps
  3. Save content to server (Task 9) - Save the article structure as a JSON file

I) SUBAGENT 5 - Integration Agent:
- Name: ArticleIntegrator
- Final Output: An HTML file containing the formatted article with text, images, and proper layout.
- Tasks:
  1. Access text files already saved (Task 10) - Load the transcription text
  2. Access text files already saved (Task 11) - Load the extracted images and article structure
  3. GPT4 (Task 1) - Combine the text and images according to the article structure, inserting the images at the appropriate locations based on the timestamps
  4. GPT4 (Task 1) - Apply formatting and styling to the article using HTML tags and CSS
  5. Save content to server (Task 9) - Save the

claude 3 - mermaid-for-all

claude 3 - subagent1

subagent1-refined

A) SUBAGENT SUMMARY: The VideoTranscriber subagent transcribes a YouTube video and generates a text file containing the transcription with timestamps for each sentence or phrase.

B) FINAL TASK OUTPUT: A text file (.txt) containing the video transcription with timestamps for each sentence or phrase, saved on the server with a URL like server/_public/timestamp/storage/transcription1.txt.

C) SUBAGENT INPUT: The URL of the YouTube video to be transcribed (e.g., https://www.youtube.com/watch?v=xyz123).

E) SUBAGENT TASK SUMMARY:
1. YouTube video URL (input) > Download youtube video to mp4 (Task 43) > MP4 file URL on internal server
2. MP4 file URL > Convert mp4 to mp3 (Task 44) > MP3 file URL on internal server
3. MP3 file URL > Get text transcription of mp3 (SRT subtitles) (Task 32) > Text transcription (SRT file) with timestamps
4. Text transcription > GPT4 (Task 1) > Clean and format the SRT file to remove unnecessary information and keep only the text and timestamps for each sentence or phrase
5. Cleaned and formatted transcription > Save content to server (Task 9) > Text file URL on the server (e.g., server/_public/timestamp/storage/transcription1.txt)

F) SILOS:
Silo 1: Video Download and Conversion
- Download youtube video to mp4 (Task 43)
- Convert mp4 to mp3 (Task 44)

Silo 2: Transcription and Formatting
- Get text transcription of mp3 (SRT subtitles) (Task 32)
- GPT4 (Task 1) to clean and format the SRT file

Silo 3: Saving Output
- Save content to server (Task 9) to save the final transcription text file

random-autocomplete

To get the end goal of a comprehensive, media-rich article based on your "how to" video, the most relevant tasks might be:

1. Transcribe the video to extract the spoken content.
2. Analyze the transcription to identify key steps and corresponding timestamps.
3. Extract relevant images from the video at the identified timestamps.
4. Organize the transcription into a structured article format with headings, paragraphs, and image placement.
5. Convert the structured content into a web-friendly format (e.g., HTML, Markdown, or a CMS-specific format).

The overall sequence would be:

1. Upload video to the Agent's virtual machine.
2. Use a speech recognition API (e.g., Google Cloud Speech-to-Text, AWS Transcribe, or Azure Speech to Text) to transcribe the video and generate a text file with the spoken content and timestamps.
3. Process the transcription file using natural language processing (NLP) techniques to identify key steps and their corresponding timestamps. This can be done using libraries like spaCy or NLTK in Python.
4. Use a video processing library (e.g., OpenCV or FFmpeg) to extract images from the video at the identified timestamps.
5. Create a structured document (e.g., using Python's built-in `dict` or a library like `pydantic`) to organize the transcription text into headings, paragraphs, and image placement based on the identified steps and timestamps.
6. Convert the structured document into a web-friendly format using a templating engine (e.g., Jinja2) or a markup language library (e.g., Python-Markdown).
7. Output the generated article in the desired format (e.g., save as an HTML file, commit to a Git repository, or upload to a CMS via an API).

By following this sequence of tasks, the AI Agent should be able to create a comprehensive, media-rich article based on your "how to" video, utilizing its access to various APIs, libraries, and the virtual machine running Windows 11.

subagent1refined-repeating

G) FULL TASKS (INCLUDING REPEATING TASKS):

1. Receive YouTube video URL as input
2. Download youtube video to mp4 (Task 43)
3. Save MP4 file URL on internal server
4. Convert mp4 to mp3 (Task 44)
5. Save MP3 file URL on internal server
6. Get text transcription of mp3 (SRT subtitles) (Task 32)
7. Save text transcription (SRT file) with timestamps
8. GPT4 (Task 1) to clean and format the SRT file
   - Remove unnecessary information
   - Keep only the text and timestamps for each sentence or phrase
9. Save cleaned and formatted transcription
10. Save content to server (Task 9)
11. Generate text file URL on the server (e.g., server/_public/timestamp/storage/transcription1.txt)
12. Output the text file URL as the final result of the subagent

In this case, there are no explicitly mentioned repetitions of the same individual tasks within the subagent. The tasks are sequential and do not require any loops or iterations based on the provided information. However, if there were specific instructions to repeat certain tasks multiple times, those would be included in the full task list accordingly.

task-ID-numbers

subagent1-output-variables

Here's the updated task list with the output tokens added:

G) FULL TASKS (INCLUDING REPEATING TASKS):

1. Receive YouTube video URL as input [youtube-url]
2. Download youtube video to mp4 (Task 43) [youtube-mp4]
3. Save MP4 file URL on internal server [mp4-url]
4. Convert mp4 to mp3 (Task 44) [mp4-mp3]
5. Save MP3 file URL on internal server [mp3-url]
6. Get text transcription of mp3 (SRT subtitles) (Task 32) [mp3-srt]
7. Save text transcription (SRT file) with timestamps [srt-file]
8. GPT4 (Task 1) to clean and format the SRT file [cleaned-srt]
   - Remove unnecessary information
   - Keep only the text and timestamps for each sentence or phrase
9. Save cleaned and formatted transcription [cleaned-transcription]
10. Save content to server (Task 9) [server-content]
11. Generate text file URL on the server (e.g., server/_public/timestamp/storage/transcription1.txt) [text-file-url]
12. Output the text file URL as the final result of the subagent [output-text-file-url]

H) OUTPUT TOKENS

[youtube-url]: The input YouTube video URL
[youtube-mp4]: The downloaded MP4 file from the YouTube video
[mp4-url]: The URL of the MP4 file saved on the internal server
[mp4-mp3]: The converted MP3 file from the MP4 file
[mp3-url]: The URL of the MP3 file saved on the internal server
[mp3-srt]: The SRT subtitles generated from the MP3 file
[srt-file]: The saved SRT file with timestamps
[cleaned-srt]: The cleaned and formatted SRT file by GPT4
[cleaned-transcription]: The saved cleaned and formatted transcription
[server-content]: The content saved to the server
[text-file-url]: The generated text file URL on the server
[output-text-file-url]: The final output text file URL as the result of the subagent

subagent1refined-repeating-json-core

A) SUBAGENT SUMMARY: A subagent that transcribes a YouTube video, cleans the transcription, and saves it as a formatted text file.

B) FINAL TASK OUTPUT: A text file containing the cleaned and formatted transcription of the YouTube video, with timestamps for each sentence or phrase. The file will be saved on a server with a generated URL (e.g., server/_public/timestamp/storage/transcription1.txt).

C) SUBAGENT INPUT: A YouTube video URL.

E) SUBAGENT TASK SUMMARY:
The subagent receives a YouTube video URL as input, downloads the video as an MP4 file, converts it to an MP3 file, generates an SRT transcription file with timestamps, cleans and formats the transcription using GPT4, and saves the final formatted transcription as a text file on a server, generating a URL for the file.

G) TASKFLOW SUMMARY:

Task 1:
Title task: Receive YouTube video URL input
Description of the task: The subagent receives a YouTube video URL as the initial input to begin the transcription process.
Input(s) involved: [YouTube video URL]
Output token name: YouTube video URL

Task 2:
Title task: Download YouTube video as MP4
Description of the task: The subagent downloads the YouTube video specified by the input URL as an MP4 file using Task 43.
Input(s) involved: [YouTube video URL]
Output token name: MP4 file URL

Task 3:
Title task: Save MP4 file URL on internal server
Description of the task: The subagent saves the downloaded MP4 file URL on an internal server for further processing.
Input(s) involved: [MP4 file URL]
Output token name: Saved MP4 file URL

Task 4:
Title task: Convert MP4 to MP3
Description of the task: The subagent converts the downloaded MP4 file to an MP3 file using Task 44.
Input(s) involved: [MP4 file URL]
Output token name: MP3 file URL

Task 5:
Title task: Save MP3 file URL on internal server
Description of the task: The subagent saves the converted MP3 file URL on an internal server for transcription.
Input(s) involved: [MP3 file URL]
Output token name: Saved MP3 file URL

Task 6:
Title task: Get text transcription of MP3 (SRT subtitles)
Description of the task: The subagent generates an SRT subtitle file containing the text transcription and timestamps of the MP3 file using Task 32.
Input(s) involved: [MP3 file URL]
Output token name: SRT transcription file

Task 7:
Title task: Save text transcription (SRT file) with timestamps
Description of the task: The subagent saves the generated SRT transcription file with timestamps for further processing.
Input(s) involved: [SRT transcription file]
Output token name: Saved SRT transcription file

Task 8:
Title task: Clean and format SRT file using GPT4
Description of the task: The subagent uses GPT4 (Task 1) to clean and format the SRT file, removing unnecessary information and keeping only the text and timestamps for each sentence or phrase.
Input(s) involved: [SRT transcription file]
Output token name: Cleaned and formatted transcription

Task 9:
Title task: Save cleaned and formatted transcription
Description of the task: The subagent saves the cleaned and formatted transcription for storage on the server.
Input(s) involved: [Cleaned and formatted transcription]
Output token name: Saved cleaned and formatted transcription

Task 10:
Title task: Save content to server
Description of the task: The subagent saves the cleaned and formatted transcription to the server using Task 9.
Input(s) involved: [Saved cleaned and formatted transcription]
Output token name: Server-saved transcription file

Task 11:
Title task: Generate text file URL on the server
Description of the task: The subagent generates a URL for the saved transcription text file on the server (e.g., server/_public/timestamp/storage/transcription1.txt).
Input(s) involved: [Server-saved transcription file]
Output token name: Transcription text file URL

Task 12:
Title task: Output the text file URL as the final result
Description of the task

subagent1mermaid

prompts-required

Based on the components mentioned above, here are the unique prompts that need to be written for this agent workflow:

Task: Video Transcription
Output of the task: A text file containing the full transcription of the "how to" video with timestamps for each sentence or phrase.
Description: Write a prompt that instructs the LLM to transcribe the given video, including accurate timestamps for each sentence or phrase. The prompt should emphasize the importance of capturing the spoken content verbatim and ensuring the timestamps are precise to facilitate image extraction and synchronization.

Task: Key Steps Identification
Output of the task: A list of the core steps involved in the "how to" process, derived from analyzing the video transcription, with corresponding timestamps or timestamp ranges.
Description: Write a prompt that guides the LLM to analyze the video transcription and identify the key steps in the "how to" process. The prompt should instruct the LLM to extract the main actions or milestones mentioned in the transcription and associate them with relevant timestamps or timestamp ranges.

Task: Image Extraction Guidance
Output of the task: A set of instructions for extracting relevant images from the video at specific timestamps mentioned in the transcription.
Description: Write a prompt that provides guidance on how to extract images from the video based on the timestamps identified in the transcription. The prompt should specify the desired image format (e.g., JPEG, PNG) and provide instructions on how to name the image files in a meaningful way.

Task: Article Structure Generation
Output of the task: A structured outline or template for the article, organizing the content into sections, headings, and paragraphs based on the identified key steps.
Description: Write a prompt that directs the LLM to generate a logical and coherent structure for the article based on the key steps identified in the "how to" process. The prompt should instruct the LLM to create an outline or template that includes sections, headings, and paragraphs, ensuring a natural flow of information.

Task: Integration and Formatting Instructions
Output of the task: A set of instructions or a script outline for combining the transcription text, extracted images, and article structure into a well-formatted article suitable for web publishing.
Description: Write a prompt that guides the development of a script or program to integrate the transcription text, extracted images, and article structure into a cohesive and well-formatted article. The prompt should provide instructions on how to place images alongside the relevant text based on timestamps and specify the desired output format (e.g., HTML, Markdown, or a CMS-specific format).

Task: Timestamp Synchronization Mechanism
Output of the task: A description of a mechanism to synchronize the text and images based on the timestamps from the transcription.
Description: Write a prompt that outlines a mechanism for synchronizing the text and images based on the timestamps from the transcription. The prompt should suggest creating a data structure or file (e.g., JSON) that maps the timestamps to the corresponding text and image elements, ensuring proper alignment of content in the final article.