We have significantly expanded the Work Analysis feature, releasing two new steps: Compare (comparing a target person's work video with the standard) and Extract (composing a model cycle from multiple cycles). Additionally, we have enhanced Cycle Analysis Support (repeat type declaration / single-cycle video support / AI model improvements / stabilization). This article introduces the details of this update.

① Work Analysis Compare Function — Visualize skill level against target person

Using the established Standard (improved video + element definitions) as reference, the system now watches and aligns repeat work videos of target persons (new hires, trainees, etc.) using AI, making it possible to visualize skill level through cycle time comparison, delay per element, and AI coaching.

• Previously: Training managers could only retrospectively say "they seem much faster now" or "this part seems slow," with no objective numbers to back it up.
• Now: By placing standard and target videos side-by-side for synchronized playback, delays per element are visualized with red shading. You can see at a glance: "You are delayed by X seconds in element YY, and you would be faster if you moved like this."

●Key features

• Select target person and video with one click to compare: Select a target person from the team member list, choose their work video from the material library, and start.
• Cycle time Compare: Display standard and target cycles in stacked bars. Includes average and standard value lines.
• Delay per element: Line graph showing difference from standard = 0. Selected cycle appears as solid line, average as dotted line.
• Variance per element: For 2+ cycles, display shortest – longest range, average, selected cycle, and standard value all on one screen.
• AI coaching: AI generates individual improvement suggestions per element based on differences between standard and target (with priority, element, key point, and advice).
• Synchronized playback of standard and target: Compare videos placed side-by-side at the same playback position.
• Cycle adjustment and re-analyze: If AI auto-splitting is misaligned, fine-tune. If unsatisfied, re-analyze.
• Re-analyze failed chunks only: If analysis fails for part of a long video, re-run just that section (no AI count consumed).

For usage details, see STEP6 Compare: Compare target person's work video with standard (skill level visualization).

② Extract Step — Compose a single model cycle from multiple cycles

From a video with repeated work, the system now visualizes cycle variation and adds a new step to build a "model cycle" to use as reference. All subsequent classification and improvement work is performed against this model cycle.

• Previously: Repeat videos could only be analyzed as "rough overall," making it difficult to combine best elements to create an ideal form.
• Now: An element-by-cycle matrix shows cycle time per element. Clicking cells—where green = fastest, red = slowest, blue ✓ = selected—makes it easy to compose a model.

●Key features

• Cycle analysis: AI automatically extracts cycles and element work from multi-cycle videos.
• Two presets:
  • [Best per element] = Bulk-adopt good takes stable per element (composite of element bests).
  • [Best single cycle] = Align all elements to the best cycle overall (consistent demonstration).
• Cell-level manual override: After preset, select individual elements from different cycles if needed.
• Model continuous preview: Play selected elements in sequence to confirm the appearance of the model on the spot.
• Cycle adjustment: Fine-tune boundaries if AI auto-splitting is misaligned.
• Single-cycle videos proceed as-is: Since single-cycle videos need no composition, just verify and proceed to classification.

For usage details, see STEP2 Extract: Extract element work and create a model single cycle.

③ Cycle Analysis Support (accuracy and stability enhancement)

To make Cycle Analysis, which supports the Extract step, even more user-friendly, we have significantly enhanced the underlying infrastructure.

●Declare presence/absence of repeat during analysis

During AI analysis of element work, the system now asks users: Does this video repeat the same work? By providing the answer as a hint to the analysis, extraction accuracy has improved.

• Yes: Analyze as multiple cycles, allowing you to align each cycle in the model step to compose the model.
• No: Analyze as one-time work, then proceed directly to classification and improvement.

●Single-cycle videos can now be analyzed directly

Previously, analysis assumed "repeat work," so videos with only one cycle sometimes had unstable results. With this improvement, single-cycle videos are now correctly analyzed as "one-time work". After verification in STEP2 (Extract), proceed directly to classification.

Step Structure Update (4 steps → 6 steps)

With the addition of new features, the Work Analysis editor has been expanded to 6 steps.

1. STEP1 Selection: Select the original video and review AI improvement suggestions.
2. STEP2 Extract (new): Extract element work + compose model cycle.
3. STEP3 Classification (formerly STEP2): Classify into 3 categories.
4. STEP4 Improve (formerly STEP3): Cut, reorder, adjust speed.
5. STEP5 Verify (formerly STEP4): Comparison preview before/after + output.
6. STEP6 Compare (new): Compare target person's work video with standard.

For overall details, see Work Analysis overview and getting started.

We look forward to your continued use of Dive.