Audio problems are the top reason viewers abandon otherwise-fine videos, yet audio is the least-reviewed part of most review rounds - reviewers watch picture with sound on in the background, and "the mix sounds fine" ships unexamined. The fix is not new software; it is a deliberate audio pass that runs inside the same review link as your picture notes: four layers, timecoded notes, and a deliverables check before final approval.
Why audio gets skipped in review
Picture bias is structural: review interfaces are built around frames, reviewers look at screens, and feedback language is visual ("that transition," "this frame"). Audio defects also hide better - a soft clip of wind noise does not look wrong, it only sounds wrong, and laptop speakers hide it entirely. The countermeasure is process, not gear: schedule one review pass where the only job is listening, on the best speakers available, and give reviewers a vocabulary for what to flag. Tools cannot replace the listening pass; they can make its notes precise.
The four-layer audio pass
| Layer | What to listen for | Typical flags |
|---|---|---|
| Dialogue | Intelligibility, level consistency between speakers and scenes | "Line at 1:24 buried under the room tone" |
| Music | Level vs dialogue, entrance/exit points, mood match | "Music swallows the CEO quote at 2:10" |
| Sync | Lip sync, sound effects landing on the action, music beat alignment | "Footsteps land two frames late at 0:47" |
| Captions + access | Caption accuracy, timing, sound-representation (or the lack of it) | "Caption says the wrong product name at 3:05" |
Run the pass layer by layer rather than once through: a dialogue-only listen, then a music-only listen (many NLEs and review tools let reviewers note moments while the editor mutes layers internally), then sync spot-checks at cuts, then the caption pass. Four short passes catch more than one distracted full watch - the caption layer especially deserves its own round, detailed in the caption review workflow.
Loudness targets worth knowing
Loudness is measurable, which makes it the one audio dimension you should treat as a spec rather than a vibe. Commonly used targets as of 2026: streaming and social platforms normalize around -14 LUFS integrated; European broadcast follows EBU R128 at -23 LUFS; US broadcast follows ATSC A/85 at -24 LKFS. Two practical rules fall out of this: ask for the delivery spec before the final mix (the destination dictates the target), and check which platform your video actually lives on - a cut normalized for YouTube then uploaded to a broadcast context misses its target by a mile. The review-side job is simpler than the mixing job: confirm the mix was made to a spec, and flag moments that sound inconsistent even if the integrated number passes. Deeper export and delivery mechanics are in what format to export for client review.
How to give timecoded audio notes
Audio notes work in the same frame-accurate comment system as picture notes - you pin to the moment where the problem occurs, which for audio is usually where it becomes audible: "boom rumble starts at 0:12," "SFX late at 1:47," "music cuts mid-phrase at 2:30." The discipline that makes them actionable: name the layer (dialogue/music/SFX), name the problem, name the desired outcome. "Audio is off at 1:47" forces a phone call; "music duck late at 1:47 so the line lands" is fixed in one pass. This is the same note precision the annotation guide covers in frame-accurate review - applied to ears instead of eyes.
Deliverables to confirm before final approval
The final-approval gate for audio should check the deliverables list, not just the sound: the mix file to spec (stereo or 5.1 as required), stems if the destination demands them (dialogue/music/SFX splits), an M&E version if international versions exist, the loudness measurement against the target, and caption files matching the final cut. Teams that confirm deliverables at approval never relitigate the mix in week six; teams that skip it meet the broadcaster's tech QC in the worst possible mood. The full pre-send checklist - audio layers included - is in the video review checklist.
What a review tool can and cannot do for audio
Honest boundaries: review tools are where audio feedback lives, not where audio work happens. Frame-accurate comments handle audio moments as precisely as picture moments; transcription surfaces support the caption layer; version stacks keep each mix iteration attached to its notes. What review tools do not do is mixing - no waveforms, no bus processing, no loudness metering inside the review link. The pass happens in the DAW or NLE; the review link is where the sign-off gets recorded. That division is exactly why the workflow above works anywhere, including Frooty's review surface.
FAQ
How do clients give feedback on audio in a review link?
The same way as picture notes: pause at the moment, pin a comment, describe what sounds wrong. Frame-accurate comments land on audio moments as precisely as visual ones. The workflow upgrade is on your side - tell clients the round includes a listening pass and give them the four layers as a vocabulary.
What loudness should social media video be mastered to?
Most streaming and social platforms normalize playback around -14 LUFS integrated, which is the practical default for social cuts. But always confirm the destination's current spec before the final mix - broadcast contexts use stricter targets (EBU R128 -23 LUFS, ATSC A/85 -24 LKFS), and the destination dictates the number, not habit.
Can I review audio separately from picture?
Yes, as a process: run a listening-only pass where reviewers focus on dialogue, then music, then sync, using timecoded comments on the same link. The review tool does not need an "audio mode" - the discipline of layer-by-layer passes over the same cut is what catches what a single distracted watch misses.
What audio files should I deliver with the final cut?
The delivery spec decides, but the common set is: the final mix to the destination's loudness spec, stems (dialogue/music/SFX) where requested, an M&E mix for international versions, the loudness measurement report, and caption files that match the final cut. Confirm the list at the approval gate - not after delivery.