Can You Fix It in Post? What Editors Can Actually Salvage in 2026

Can You Fix It in Post? What Editors Can Actually Salvage in 2026

The Most Expensive Four Words on a Shoot

"Fix it in post" is the cheapest sentence to say while the camera is rolling and the most expensive one to act on afterwards. Across the 10,000+ projects our team has cut, roughly one in six arrives with a capture defect the client assumed the edit would quietly absorb. Most of them we can absorb. A small, entirely predictable set we cannot — and no budget, no plugin, and none of the 2026 AI restoration tools change that, because the information was never recorded in the first place.

That is the entire rule, and it is worth memorising before your next shoot: post-production redistributes information. It cannot invent it. Noise reduction moves noise around. A grade redistributes contrast. Upscaling redistributes existing edges into more pixels. Every one of those tools is working with what the sensor and the microphone actually captured. When a channel is empty, it stays empty.

The Salvage Table

Here is how we triage incoming footage, in the order we check it. "Expensive" means recoverable but billable — extra hours, extra passes, and usually a visible compromise somewhere else in the frame.

Problem on the sourceVerdictWhat it actually costs
Wrong white balance, flat log profileFixableMinutes, if you shot 10-bit
Room hiss, AC hum, laptop fanFixableOne processing pass
Framing slightly looseFixableFree, if you shot 4K and deliver 1080p
Handheld shakeExpensiveStabilisation crop of 5–15% plus rolling-shutter cleanup
Underexposed by 2+ stopsExpensiveNoise reduction, softened detail, grain rebuild
Mixed colour temperature in one frameExpensiveManual masking, shot by shot
Blown highlightsPermanentNothing to recover — the channel is already at maximum
Out-of-focus footagePermanentReshoot
Clipped, distorted audioPermanentRe-record
No coverage, no second anglePermanentReshoot
Every capture defect lands in one of three buckets: fixable in a single pass, expensive but recoverable with a visible compromise, or permanent — meaning the only fix is to shoot it again.
Every capture defect lands in one of three buckets: fixable in a single pass, expensive but recoverable with a visible compromise, or permanent — meaning the only fix is to shoot it again.

The pattern is consistent. Anything that is a distribution problem — the values exist but sit in the wrong place — is fixable. Anything that is a missing data problem is not. Over 200M+ views' worth of delivered work and $10M+ in client revenue generated, we have never once made an out-of-focus shot sharp or an overdriven microphone clean. We have gotten very good at hiding both, which is a different service and a slower one.

Audio Is Where AI Genuinely Moved the Line

If you last formed an opinion about audio repair in 2022, update it. Speech-enhancement models — Adobe Premiere Pro, Descript's Studio Sound, and Voice Isolation inside DaVinci Resolve — now do in one click what used to be a specialist's afternoon. A camera-mic interview recorded in a hard-surfaced room is genuinely rescuable today.

What those models fix: broadband hiss, electrical hum, air conditioning, keyboard noise, and a meaningful amount of room reverb.

What they still cannot fix, and will not in 2027 either:

  • Clipped audio. When the preamp drives past 0 dBFS, the waveform's peaks are flattened. Declipping tools interpolate a guess at the missing shape; they do not restore it. On sustained speech, that guess sounds like a robot.
  • Two people on one microphone talking simultaneously. Separation models improve every year and still leave artefacts on overlapping speech.
  • A voice that was never in the recording. Directional mic pointed the wrong way, wireless dropout, dead battery. There is nothing to enhance.
Speech-enhancement models genuinely fix broadband noise and electrical hum in a single pass. Clipped audio and out-of-focus footage remain unrecoverable, because in both cases the information was never captured.
Speech-enhancement models genuinely fix broadband noise and electrical hum in a single pass. Clipped audio and out-of-focus footage remain unrecoverable, because in both cases the information was never captured.

Set your levels to peak around −12 dBFS and this entire category of problem disappears. We go deeper on capture chains and mixing in our sound design and music selection guide.

Picture: The Upscaling Myth

Topaz Video AI and its competitors are excellent tools that are widely misunderstood. Upscaling a 1080p clip to 4K produces a plausible, sharper-looking image by generating detail that was never in the source. That is fine for archival footage and legitimately useful for old material. It is not a substitute for shooting at delivery resolution, and it does not un-blur anything, because focus error destroys the edge information the model would need to reason from.

Three picture rules we hold every client to:

  1. Shoot at least one resolution above delivery. 4K acquisition for a 1080p deliverable buys you reframing, stabilisation crop, and punch-ins for free. YouTube's recommended upload encoding settings reward the extra data on the way out, too.
  2. Protect the highlights, not the shadows. A blown highlight is at the sensor's ceiling and has no data left. Underexposure, in 10-bit log, is usually recoverable. When in doubt, expose for the sky.
  3. Shoot 10-bit if the camera offers it. It is the single biggest difference between a grade that holds up and one that bands. Our colour grading guide covers what that buys you shot by shot.

The 60-Second Pre-Roll Gate

Nearly every permanent defect we see would have been caught by one minute of checking before the take. Run this every single time you press record — not once at the start of the day.

The 60-second pre-roll gate: check framing, confirm focus on the eyes, watch the audio meters during a real sentence, then verify card space and battery. One minute here prevents almost every permanent defect.
The 60-second pre-roll gate: check framing, confirm focus on the eyes, watch the audio meters during a real sentence, then verify card space and battery. One minute here prevents almost every permanent defect.
  1. Framing — headroom correct, and nothing important sitting where captions will land
  2. Focus — punch in on the eyes at 100% and confirm, don't trust the small screen
  3. Audio — say a full sentence at performance volume and watch the meters peak near −12 dBFS
  4. Space — card space and battery confirmed for the full take, plus a margin
  5. Frame rate — set for the intended use now; you cannot create 120fps slow motion from a 30fps clip
  6. Backup — dual-card recording on, or a phone recording audio as a safety

That last one matters more than the other five combined. A second audio source costs nothing and has saved more of our client projects than any plugin. Where that file goes afterwards is covered in our video file management system.

What to Do When the Footage Is Already Broken

Assume you are reading this after the shoot. Your options, in order:

  • Reframe around it. A soft wide can become a cutaway; a clipped word can be cut on the b-roll.
  • Replace, don't repair. Bad on-camera audio plus a clean phone recording of the same take is a better result than any enhancement.
  • Cover it. Broll, a graphic, or a cut to another angle beats a heroic restoration nobody notices anyway.
  • Rewrite around the gap. If a section is unusable, a voiceover line often bridges it more cleanly than a damaged clip.
  • Reshoot the one shot. Usually cheaper than the six billable hours of salvage, and always better looking.

Tell your editor which of these you want before they start. Our editor brief handoff has the exact structure for flagging known problem clips so nobody spends an afternoon on a shot you were happy to cut.

The Bottom Line

Post-production can move information around; it cannot create information that was never captured. Focus, clipped audio, blown highlights, and missing coverage are permanent — everything else is a question of how many billable hours you want to spend and what you are willing to compromise to get it.

Sixty seconds of checking before each take is the highest-return minute in video production. Nothing our team does in the edit bay comes close to it.

👉 Start Your Project Now

Frequently asked questions

Can you fix out-of-focus video in post?
No. Focus error destroys the edge detail that sharpening and AI upscaling tools need in order to reconstruct an image. Those tools generate plausible new detail rather than recovering what was lost, which on a soft face reads as waxy and artificial. The only real fix for out-of-focus footage is a reshoot, or cutting to a different angle.
Can clipped or distorted audio be fixed?
No, not properly. When a recording level pushes past 0 dBFS the peaks of the waveform are flattened and permanently discarded. Declipping tools interpolate a guess at the missing shape, which sounds acceptable on a single short word and robotic across sustained speech. Record with peaks around minus 12 dBFS and keep a second audio source as a safety.
What is the difference between underexposed and blown-out footage?
Underexposure leaves the image data intact but sitting low in the signal, so lifting it in a grade usually works at the cost of visible noise. Blown highlights are already at the sensor's maximum value, meaning no data exists to recover at all. Underexposure is expensive to fix; blown highlights are permanent. When in doubt, expose to protect the highlights.
Does AI video upscaling actually work?
It works for what it is designed to do, which is generating a plausible higher-resolution image from existing detail. Tools like Topaz Video AI are genuinely useful for archival and older footage. They do not restore lost information, do not sharpen out-of-focus shots, and are not a substitute for acquiring at least one resolution above your delivery format.
Why do editors ask you to shoot in 4K for a 1080p video?
Because the extra resolution becomes editing headroom. Delivering 1080p from a 4K source allows reframing, punch-ins for a second angle, and the 5 to 15 percent crop that stabilisation consumes, all without any quality loss. Shooting at delivery resolution means every one of those adjustments costs visible sharpness.
What should I check before every take?
Five things, in about a minute: framing with correct headroom, focus confirmed by punching in on the eyes at 100 percent, audio meters peaking near minus 12 dBFS during a full spoken sentence, card space and battery for the whole take, and the frame rate set for its intended use. Add a phone recording as a backup audio source.
Is it cheaper to reshoot or to fix footage in post?
Reshooting a single shot is usually cheaper. Salvage work on a badly captured clip routinely runs several billable hours of noise reduction, masking, and stabilisation, and still ends in a visible compromise such as a softer image or a tighter crop. Reshoot when the setup still exists; cover or cut around it when it does not.