Video summary

اصنع صورك المصغرة بالذكاء الاصطناعي بدون فوتوشوب!

Main summary

Key takeaways

Technology

Overview

The tutorial explains how to create YouTube thumbnails using AI instead of Photoshop. It emphasizes that success is not just generating an attractive image, but producing a thumbnail that is:

  • Professional
  • Relevant to the video content
  • Click-attracting (creates curiosity)

Core Workflow: AI thumbnail creation with script + prompting

The video uses an AI image generator (referred to as “ShadGBT / ChatGPT”, i.e., ChatGPT with image capabilities). The key idea is to provide the model with context from the video script, so it can connect:

  • the video outcome
  • the main problem/challenge
  • the relationship between video title and thumbnail
  • the viewer emotion/curiosity you want to trigger

Script extraction (helper step)

To get the full context, the creator suggests extracting the full script from YouTube using a browser extension mentioned as “YouTube SummerD” (full name/link is requested in the comments).

Method 1: “Quick method” (less context → weaker optimization)

  1. Give the AI a simple instruction, such as: “Design a thumbnail about AI video production.”
  2. The AI may produce a decent thumbnail, but the creator warns it likely won’t be the best because it lacks:
    • video specifics
    • the hook
    • the main outcome
    • the intended viewer feeling

Method 2: “Professional method” (script + structured prompt → better results)

The tutorial demonstrates this method with two experiments.

Experiment 1: Replicating a top-channel style (e.g., “Mr. Best”)

  1. Choose a target video (example: a “trapped in the sky for 30 days” challenge).
  2. Extract the script using the YouTube SummerD extension.
  3. Copy the script text.
  4. Send it to ChatGPT with a prompt to:
    • analyze as a YouTube expert
    • identify the outcome, biggest challenge, and problem
    • generate five thumbnail concept ideas using elements like:
      • outcome
      • problem
      • comparison
      • curiosity
    • choose the strongest concept
  5. Generate an image using constraints such as:
    • 16:9 dimensions
    • no text
    • include a selected character image (e.g., a Mr. Beast image provided as input)
  6. The creator notes the result resembles Mr. Beast thumbnail style and can be edited, for example:
    • “Change clothes to red”
    • “Add more details”

Feature highlighted

  • Iterative control: edit the generated image by requesting specific modifications.

Experiment 2: Creator’s workflow with more control + localization fix

  1. Pick the creator’s video (about creating “Dark History” long videos using AI).
  2. Extract the script (optionally in Arabic if the video is Arabic).
  3. Provide ChatGPT with:
    • a “basic promo” template (the creator offers to share it via comments)
    • the full script
  4. ChatGPT analyzes:
    • strongest content points
    • best visual element
    • the problem/solution connection
    • how the title connects to the thumbnail
  5. Generate a thumbnail using attached images (e.g., snippets/screens used in the video).
  6. First attempt: the output became “scene-like,” with 102 scenes, and included errors (e.g., language should be Arabic).
  7. Second attempt: improve results by using a more detailed prompt, including layout direction such as:
    • background matching the “Dark History” tone and colors
    • three video screens on the right with screenshots from the provided images
    • professional text on the left (examples mentioned: “Best Content,” “Dark History,” “Free,” “Create it Yourself”)
    • professional, eye-catching, intriguing tone
    • enforce correct 16:9 aspect ratio
  8. The final results were described as better because the prompt became more explicit about structure and composition.

Feature highlighted

  • Prompt engineering for precise layout (multi-panel composition + text placement + style constraints).

Editing capability: generate variations on demand

After generating a thumbnail, the creator demonstrates you can request changes like:

  • a different style
  • adding/changing/removing text
  • switching backgrounds (e.g., white background)
  • inserting the creator’s own photo by:
    • uploading the person’s image
    • instructing AI to place it as a main element with correct sizing and positioning

Key Takeaway

AI thumbnail generation performs best when you provide:

  • script/context
  • structured prompts that control:
    • concept selection
    • composition/layout
    • style and color theme
    • localization (language)
    • aspect ratio (16:9)
    • image inputs (characters, screenshots, personal photos)

Main speakers or sources

  • Speaker/creator: Mahmoud (referenced directly)
  • Primary AI system used: ChatGPT / “ShadGBT”
  • Script extraction tool: YouTube extension “YouTube SummerD” (name may be approximate due to subtitles)
  • Example channel referenced: “Mr. Best”

Original video