AI Semantic Clipping
AI analyzes long-form videos and suggests clip-worthy scenes for extraction.
Overview
Sovran's AI can analyze long-form videos — testimonials, podcasts, product demos — and automatically identify segments worth extracting as standalone clips. Each candidate comes with a confidence score so you can quickly review and approve the best moments.
Running Analysis
- Navigate to any video asset in your Media Library and open the detail view.
- Go to the Clips tab.
- Click "Analyze Now" to start the analysis.
The AI runs both transcript-based and visual analysis. Results appear automatically when processing completes.
Assets must be at least 4 seconds long to run analysis.
Script vs. Visual Candidates
The AI generates two types of candidates:
- Script candidates — Detected from transcript analysis. The AI identifies 15 module types: Hook, Problem, Solution, Benefit, Results, Social Proof, CTA, and more. Each segment is tagged with its module type and a confidence score.
- Visual candidates — Detected from visual scene analysis. The AI identifies visually compelling moments like product close-ups, scene transitions, or on-screen text.
Each candidate displays its timestamp range, label, description, and confidence score (0–100%).
Reviewing Candidates
- Approve — Mark a candidate as ready for export.
- Reject — Hide a candidate from the main view. Rejected candidates can be restored using the "Show rejected candidates" toggle.
- Approve All — Bulk-approve all pending candidates at once.
The info bar shows counts for total, approved, exported, and rejected candidates.
Trimming Before Export
Click "Trim" on any candidate to fine-tune the start and end points before exporting. The trimmer modal lets you adjust the boundaries while previewing the clip.
You can also toggle "Include audio" on or off for all exports.
Exporting Clips
- Single export — Click "Export" on an approved candidate.
- Bulk export — Click "Export All Approved" to export every approved candidate at once.
Exported clips are rendered server-side and appear as new assets in your Media Library. Track export progress with the status badges: Queued → Rendering → Ready.
Once a clip is ready, click "Open clip" to preview or use it in your sequences and video editor.