We've completely overhauled the AI processing for generating SOPs from videos by introducing a new large-scale model called VLLM.

This enables AI to understand the "meaning" of video content when generating SOPs, with significant improvements in analysis speed and support for longer videos.

① VLLM AI Processing for Videos — Enabling AI to Understand Video "Meaning"

Previous AI processing for motion analysis focused mainly on "dividing" the video. While scene transitions could be detected, the processing couldn't account for what the video was actually showing.

With the newly introduced VLLM (a large-scale AI that understands both video and language), the AI can now understand the actual content of the footage and generate SOPs that follow the proper context.

● Specific improvements

• Can now understand the "meaning" of videos
  Previous: Could only divide the video
  Current: Can split steps based on what action is being performed in each scene
• Can now create Subtitles from Actions
  Previous: Only transcription of spoken audio
  Current: Auto-generates Subtitles describing action content even in scenes with no dialogue
• Nonsensical transcriptions have been drastically reduced
  Since transcription now accounts for the overall video context, incoherent Subtitles caused by mishearing have been greatly reduced
• Auto-generation of Questions, Text, and Shapes
  When exporting SOPs from videos, the AI now automatically inserts verification Questions and Text Shapes (annotations) within the video

② Improved Analysis Speed and Extended Video Length Support

We've completely refreshed the processing infrastructure to deliver speeds and capacity that work well on the shop floor.

• Analysis speed has been greatly improved
  Processing wait time after Upload has been significantly reduced
• File splitting every 10 minutes has been eliminated. Now supports up to 50 minutes per File.
  Previously, videos over 10 minutes were automatically split by the System, but now a single video can be processed as-is up to a maximum of 50 minutes. Training videos or full operation workflows that run longer can now be converted to SOPs without splitting.

Existing Created video SOPs will not be affected. The new AI applies when you Upload a new video or when you Execute "AI Re-Process" on existing videos.

Thank you for your continued use of Dive◎