How does subject analysis work?

Subject analysis is Smooth Operator's way of understanding who and what is in your video. It detects faces, people, and bodies in your footage so that the AI can make smarter camera movement suggestions.

What subject analysis does

When you use the AI Camera Assistant for the first time in a project, the app automatically analyzes your video to:

• Detect faces and people in each frame.
• Track subjects across multiple frames to understand where they move.
• Build a map of where the important subjects are at each point in time.

This information helps the AI generate camera movements that keep subjects properly framed.

How to see subject analysis in action

1. Open the AI Camera Assistant in any project.
2. Send your first message.
3. Before the AI responds, you will see a status bar:
   • "Analyzing video for faces/bodies... X%" with a purple progress bar.
4. When complete, one of three results appears:
   • "Found X subject(s)" (green) — Subjects were detected successfully.
   • "No faces/bodies detected in video" (orange) — The analysis ran but found nothing.
   • "Subject tracking failed: [error]" (red) — The analysis encountered an error.

What gets detected

Subject analysis can identify:

• Faces — Individual faces in the frame.
• People — Full-body detections.
• Objects — General object detection (less common).

Each detected subject is labeled (e.g., "Face 1," "Person 1") and tracked across frames.

Analysis phases

The analysis goes through several phases internally:

1. Loading video — Preparing the video for analysis.
2. Analyzing frames — Processing each frame (e.g., "Analyzing frame 45/300").
3. Tracking subjects — Following detected subjects across frames.
4. Building tracks — Creating continuous subject tracks.

On Apple Silicon Macs, the analysis can use the Neural Engine for faster processing.

What happens if analysis fails

Subject analysis is best-effort — it is not required for AI features to work. If it fails:

• The AI Camera Assistant still works, but its suggestions may be less precise about subject placement.
• The Video Analysis Plan continues without subject data.
• You may see "No faces/bodies detected in video" if your footage does not contain recognizable subjects.

Performance and privacy

• Analysis runs entirely on your device using Apple's Vision framework. No video data is sent to external servers for subject detection.
• Analysis time depends on video length, resolution, and your device's processing power.
• The results are cached in your project so they do not need to be re-analyzed each time.

Tips

• Well-lit, front-facing subjects are detected most reliably.
• Crowded scenes may produce many detected subjects — the AI prioritizes the most prominent ones.
• Very dark or abstract footage may not produce useful subject data. This is expected.
• Subject analysis runs automatically on your first AI message — you do not need to trigger it manually.

Related articles

• How do I use the AI Camera Assistant?
• How do I generate a Video Analysis Plan?
• What is the difference between AI Camera Assistant and Video Analysis Plan?