This release focuses on video AI analysis improvements. Beyond letting AI handle splitting automatically, you can now specify your target step granularity and align splits with existing SOP step breaks, so AI analysis matches your workflow intent. We've also improved split accuracy for long videos and made re-analysis easier to use.

① Added "Target step count" to AI processing options (Video)

Until now, step granularity was only adjustable via [Split granularity] (Process / Step / Action / Single), so fine-tuning like "split a bit finer" or "keep this one video to about 10 steps" wasn't possible. We've now added [Target step count] to the [Advanced settings] in AI processing options.

• Auto (default): AI judges granularity based on the split granularity setting, as before
• Specify step count: Set roughly how many steps to divide the entire video into (e.g., 10 steps)
• Specify seconds: Set the target length per step (e.g., 30 seconds/step)

When you specify a target, the way steps are split (breaking at meaningful work boundaries) doesn't change. We adjust only the count to match your target. Long videos are analyzed internally in sections, but adjustment targets the entire video, so step counts stay consistent regardless of video length.

● Caution

• If you don't change the setting, analysis results remain as before (Auto is the default)
• It won't display when split granularity is [Single] (treating the entire video as one step)
• When aligning to document step breaks using [Reference SOP] (described below), the document's step count takes priority, so target step count is not used (the display shows Disabled)
• This is a target, not a guarantee of hitting that exact number

● AI processing options > Advanced settings > Target step count

② Improved reference SOP step division (beta)

We've improved the feature (currently in beta) that lets you point to an existing SOP (PDF / Excel / Word / image) in [Reference SOP] so the video splits align with your document's step breaks.

• Long videos now match document step counts: Previously, long videos would have misaligned breaks, and the resulting SOP steps wouldn't match your document. Now we align the entire video to your document's step count and order
• Steps not shown in video no longer drop out: If a step was in your document but not filmed, it would disappear from the final SOP. Now we keep all document steps and assign video footage to steps where it exists
• Clearer checkbox label: [Strictly align video split to this document's steps] → [Split according to the document's steps]

See Specify a reference SOP in video analysis (beta) for usage. Since this is still beta, please share any successes or unexpected splits at customer-success@episotech.com (Dive Customer Success).

③ Re-analysis now inherits previous analysis conditions

When re-running [Re-run AI] on materials or retrying failed segments, the previous settings wouldn't carry forward, so configured settings wouldn't actually apply. Now we inherit the following settings:

• Reference SOP and how to use it (split according to document steps / use only as reference for terms and context)
• Target step count
• Silent cut setting

When retrying only some sections of a long video, they're now analyzed under the same conditions as other sections, so you won't see inconsistent granularity or style between sections.

Related information

• AI processing options: each item and how to choose
• Specify a reference SOP in video analysis (beta)
• Run AI analysis on videos that haven't been analyzed (apply AI processing later)
• How to re-run failed AI processing

For questions about this update, feel free to contact customer-success@episotech.com.