Enhancement Request: Slide Transition Detection for Loom Videos

Problem Statement

Currently, the action focuses on analyzing GitHub issue videos. However, there's a valuable use case for analyzing longer-form educational content like videos containing slide deck presentations. Users want to:

• Identify when slides change in hour+ long knowledge-sharing sessions
• Get timestamps for the first 2 minutes of each slide for efficient learning

Proposed Solution

Create a new workflow and script, based on

https://github.com/githubnext/action-genai-video-issue-analyzer/blob/main/.github/workflows/genai-video-issue-analyzer.yml
https://github.com/githubnext/action-genai-video-issue-analyzer/blob/main/genaisrc/action-video-issue-analyzer.genai.mts

Call it video-slide-deck-annotator instead of video-issue-analyzer .

0. Load the file from git storage in the repo

1. Visual Analysis Enhancement

   • Detect significant visual changes that indicate slide transitions
   • Use computer vision to identify when slide content changes substantially
   • Filter out minor changes (cursor movement, highlighting) vs major transitions

2. Timestamp Generation

   • Generate a list of timestamps where slide transitions occur
   • Include confidence scores for each detected transition
   • Export results in multiple formats (JSON, CSV, markdown)

3. Sample Detection Output

   {
     "video_duration": "01:23:45",
     "slide_transitions": [
       {
         "timestamp": "00:02:15",
         "confidence": 0.95,
         "slide_number": 1,
         "description": "Title slide to agenda"
       },
       {
         "timestamp": "00:05:30", 
         "confidence": 0.88,
         "slide_number": 2,
         "description": "Agenda to introduction"
       }
     ],
     "recommended_segments": [
       {
         "start": "00:00:00",
         "end": "00:02:00",
         "slide": 1,
         "description": "First 2 minutes of title slide"
       }
     ]
   }