How to Fix Eye Contact in Final Cut Pro
Final Cut Pro has no eye contact correction. The 12.x releases added a run of intelligent features — Generate Captions, Auto Mask, Match Color, Edit Detection, transcript search — but nothing that redirects your gaze toward the lens. We also found no FxPlug effect or workflow extension that does it at the time of writing, which is unusual for an editor with such a deep plugin market.
So if you read a script on camera and edit in Final Cut, gaze correction happens outside the app, and the job is getting the corrected file back into your library without breaking an edit. This guide covers the workflows that hold up, the export settings that matter, and what it costs for a channel that publishes every week.
What Final Cut can do on its own
For mild drift, Final Cut's own tools go a surprisingly long way. Punching in slightly and recentring the frame reduces how far off-axis your gaze reads. Trimming to the takes where you were closest to the lens helps, and the Flow transition can smooth the jump cuts that leaves on a static talking head.
None of that changes where your pupils point. If your script sat on a monitor beside the camera, every take carries the same sideways gaze, and you need a correction tool to fix it.
The options that fit a Final Cut workflow
- Correct the file before it enters Final Cut — a browser tool like GazeFix takes your talking-head file and returns a corrected MP4 you import as normal. There's no plugin to install or keep updated across Final Cut releases.
- Round-trip through another editor — Descript, VEED, Captions and CapCut all correct eye contact inside their own apps. It works, but you're importing into a second editor and exporting again, which adds time and another round of compression.
- Live correction — NVIDIA Broadcast and the Eye Contact setting in FaceTime on iPhone correct a live camera feed. They're built for calls, not for files in a library, and NVIDIA Broadcast needs a Windows PC with an RTX card anyway.
Workflow 1: correct before you import
The cleanest routine is to correct the raw talking-head recording first and import only the corrected file. Nothing in your project ever references the uncorrected media, so there's nothing to relink later and no risk of a mismatch.
Keep corrected clips in their own keyword collection, or add a Corrected keyword on import, so you can tell at a glance which footage is done. B-roll, screen recordings and anything without a face never needs correcting — leaving it out is where most of the cost saving comes from.
Workflow 2: already edited? Relink the clip
If you've already cut the video, Final Cut's File > Relink Files can point a clip at new media while keeping your edits. Final Cut checks that the replacement is compatible — among other things, a matching frame rate and a similar duration — and refuses files that aren't.
GazeFix keeps the duration and audio of your upload, but a few things change on the way through, and each one decides whether a relink is accepted and stays in sync:
- Frame rate: constant-frame-rate camera footage keeps its rate. iPhone footage recorded at a variable frame rate comes back at a constant rate of at most 60fps — check sync at the end of the clip, or use Workflow 1 for phone footage.
- Resolution: output is watermark-free at up to 1080p, so 4K media comes back at 1080p and will be scaled up in a 4K project, losing sharpness. For 4K deliveries, patch only the talking segments instead.
- HDR: iPhones record HLG HDR by default. Corrected files come back in standard dynamic range (Rec. 709), so in an HDR project they won't match neighbouring clips. Turning HDR video off on the phone avoids the problem entirely.
- Codec: output is H.264 with AAC audio. If you shot ProRes, do heavy grading after correction, not before, because H.264 leaves less latitude.
- Timecode: if your camera embedded timecode, the corrected MP4 may not carry it. Check alignment after relinking anything that was synced by timecode.
Workflow 3: patch only the parts that need it
Usually only part of a video has a gaze problem — the sponsor read, the section where you read a list of figures. Set a range around it, use Share > Export File with H.264 at your project's frame rate, correct that file, then place the result as a connected clip above the original, aligned to the same frame. Disable or lower the original's audio so they don't double up, and you can toggle the patch off to compare.
Every upload costs at least one credit, so several short patches uploaded one by one add up. Put them back to back in a scratch project, export once, correct once, and cut them back in: five twenty-second patches then cost two credits instead of five.
Export settings that avoid wasted uploads
- H.264 in an MP4 or MOV, at your project's frame rate.
- 1080p or smaller — larger files are scaled to 1080p anyway, so exporting bigger only slows the upload.
- Standard dynamic range. If you shoot log, apply your conversion LUT before exporting; flat log footage is a harder input. Test a short clip first.
- Under 500 MB, the upload limit. At a typical 10–16 Mbps that's roughly four to six minutes of 1080p — split longer sections.
- Your face reasonably large in frame. In a wide shot the eyes are too few pixels for a convincing correction.
What it costs for a weekly channel
One credit corrects one minute, rounded up per upload. Correcting six minutes of talking head in each weekly video is 312 credits a year. The Pro pack is 200 credits for $79 (about $0.40 a minute) and Studio is 600 credits for $199 (about $0.33). Credits never expire, so a fortnight off costs nothing and there's no plan to cancel.
If you'd otherwise subscribe to Descript or Captions only for eye contact, compare a year of that plan against the minutes you actually correct. If you already pay for one of them and edit there happily, its built-in correction is the better deal — the point of a file-based tool is not having to change editors.
Better input, better result
Correction is convincing over small angles and short glances; across long stretches of reading from a monitor far off to one side, it can look uncanny. Putting the script directly under or behind the lens, keeping the text narrow, and recording in short chunks improves every result. Our guide to reading a script on camera covers the setup.
Test on your worst minute, not your best. GazeFix gives you one free credit on sign-up with no card required — enough for a full minute. Export the stretch where you know you were reading, correct it, drop it above the original as a connected clip and toggle between them.