Video Guide Maker

Turn lectures into study-ready guides.

Upload a video and its transcript. Get a WCAG-ready HTML guide with the cleanest slide frames, aligned narration, and — optionally — Claude-generated section titles, key terms, alt-text, and equations you can review and refine before publishing.

Generating your study guide…

starting Uploading… 


      
    


    


      

        
Generate study guide


        

          
1. Upload


          

            
            
          


          

            
            
          


          

            
            
          

        


        

          
2. AI assist (optional)

          

            Paste an Anthropic API key to let Claude generate the fields below. Without a key the
            pipeline still runs — you'll just get the OCR-derived scaffold with placeholders.
            Toggle off any feature you don't want to pay for.
          


          

            
            
            Used once for this run. Sent over HTTPS, held in memory only, never written to disk or logs.
          


          

            
            
            
            
            
          

        


        

          
3. Media extras

          

            Local-only enrichments — no API key needed.
          

          

            
          

        


        

          
4. Output

          

            
            
            
          

        


        

          

            Advanced settings
            


              

                

                  
                  27
                

                
                
More scenesFewer scenes

                
                  How visually different two frames must be before they count as a new scene.
                  Lower (5–15) catches every animation step; higher (35+) only major slide changes.
                
              


              

                

                  
                  off
                

                
                
Off60 s

                
                  Drop new scenes that arrive less than this many seconds after the previous kept one.
                  Useful for slides that build up gradually — try ~5 s to keep just the final state.
                
              


              

                

                  
                  unlimited
                

                
                
                
1200

                
                  Hard limit on the number of frames in the guide. Useful for very long lectures
                  or to bound LLM extraction cost. Frames are evenly distributed across the video.
                
              


              

                

                  
                  0.12
                

                
                
Strict (drop more)Lenient (keep more)

                
                  Fraction of the frame a detected face must occupy before we drop it as
                  instructor-only. Lower = more aggressive; raise it if your slides have an
                  inset webcam you want to keep.
                
              


              

                
                
                  Faster; on-screen text and the heuristic LaTeX wrapping in the guide will be empty.
                  Doesn't affect LLM extraction (which reads the frame image directly).
                
              


              

                
                
                  Halves OCR time and avoids the inverted pass producing garbled near-duplicates of body text.
                  Trade-off: white-on-coloured callout text won't be recovered. Leave off for slides with coloured callouts.
                
              


              

                

                  
                  en
                

                
                
                  BCP-47 code (en, en-US, fr…).
                  Sets the lang attribute on the generated HTML and selects
                  Tesseract language packs when available.
                
              


            

          

        


        

          Processing usually takes ~30 s per minute of video.