From Shoot to Selects in Minutes: Searching Multicam Interview Footage

You just wrapped a two-day interview shoot. Four subjects, two cameras per subject, eight hours of footage total. The producer needs selects by end of day. Every quote about "company culture" and "leadership transition." Plus any reaction shots of the CEO.

Traditionally you sit down with the footage, scrub through all eight hours, take notes, mark selects, and build a timeline by hand. That is a full day of work, minimum.

The scenario

• Footage: 8 hours across 8 camera cards (2 cameras per subject, 4 subjects)
• Format: ProRes 422 HQ from two Sony FX6 cameras
• Deliverable: Selects of every mention of "company culture" and "leadership transition," plus reaction shots of the CEO
• Deadline: End of day

Step 1: Import and process

Drop all eight camera cards into FrameQuery as source folders. Start processing.

FrameQuery extracts sampled frames and audio from each clip on your device and sends only those for analysis. Your originals never leave your machine, and the extracted data is discarded the moment analysis completes. The processing pipeline runs in parallel across all clips:

• Transcription extracts every word spoken, with timestamps
• Face detection identifies and clusters every face across all cameras and subjects
• Speaker diarization identifies who is speaking when
• Scene descriptions generate natural-language descriptions of what is happening visually

For eight hours of ProRes footage, processing takes a fraction of the footage duration. You can start searching as clips finish indexing.

Step 2: Search the transcript

First request: every mention of "company culture."

Type "company culture" in FrameQuery's search bar. Results come back instantly, showing every moment across all eight hours where someone discusses company culture. Semantically related mentions come back too: "how we work together," "the environment here," "what makes this place different."

Each result shows:

• The clip and camera it came from
• The exact timecode
• A preview of the surrounding transcript
• A thumbnail of the frame

Click any result to preview the moment in context. You see and hear exactly what was said, with a few seconds of lead-in and lead-out.

Repeat for "leadership transition." Same process, new results. Every discussion of leadership changes, succession planning, new direction, across all subjects and cameras.

Step 3: Find the CEO's reactions

The producer also wants reaction shots of the CEO.

FrameQuery has already clustered faces across all your footage. Find one clip where the CEO appears, identify the cluster, and now you can search for every frame where the CEO is on screen.

Filter for clips where the CEO is visible but not speaking. These are your reaction shots: the CEO listening, nodding, reacting to what other subjects are saying. Because you shot multicam, the B camera was catching these reactions while the A camera covered the speaker.

Step 4: Review and select

You now have three sets of results:

1. Every mention of "company culture" (both cameras, all subjects)
2. Every mention of "leadership transition" (both cameras, all subjects)
3. CEO reaction shots (B camera, across all interviews)

Scrub through the results. Mark the ones you want. Reject the ones that do not work. You make these editorial decisions while reviewing a set of relevant moments instead of while scrubbing eight hours of footage.

Step 5: Export to your timeline

Select your marked clips. Export as FCPXML.

Open the FCPXML in DaVinci Resolve (or Final Cut Pro, or your NLE of choice). Your selects appear on a timeline, already trimmed to the relevant sections. The source media links automatically because the clips point to the original files on your drives.

The culture quotes are on one track. The leadership quotes are on another. The CEO reactions are on a third. Your editor can start assembling the rough cut immediately.

What this saves

Traditional workflow:

• Watch 8 hours of footage at 1.5x speed: ~5 hours
• Take notes and mark timecodes: done while watching, but adds cognitive load
• Manually locate and import selects into NLE: ~1 hour
• Total: ~6 hours minimum, and that assumes you do not miss anything on the first pass

FrameQuery workflow:

• Process footage: runs in the background, does not require your attention
• Search for "company culture": 30 seconds
• Search for "leadership transition": 30 seconds
• Find CEO reaction shots: 1 minute
• Review and mark selects: 30-45 minutes (depends on volume of results)
• Export FCPXML and import to timeline: 2 minutes
• Total: ~45 minutes of active work

A full day of scrubbing becomes under an hour of editorial decisions.

Where this workflow fits

Multicam interviews combine several properties that suit this workflow:

• Multiple cameras mean multiple angles of the same moment. Face recognition and speaker diarization keep track of who is on which camera.
• Long-form content means hours of footage where the relevant moments are scattered. Transcript search finds them all without scrubbing.
• Specific deliverables ("every mention of X") translate directly into search queries.
• Tight deadlines make the time savings most valuable.

The same approach works for corporate video production, documentary interviews, podcast recordings with video, panel discussions, and any long-form talking-head content where you need to find specific moments quickly.

The compound effect

Every interview project you process gets added to your searchable archive. Next quarter, when the client wants to revisit the "leadership transition" topic, you search your entire archive and find every relevant moment across every project you have shot.

Your interview footage becomes a searchable library of every conversation you have recorded.

---

Download FrameQuery to try multicam interview search with your own footage.