Professional video editing

What actually happens in post-production

Editing runs through six stages in a fixed order, and skipping one costs time later. Assembly, rough cut, fine cut, picture lock, finishing, delivery. Each stage answers a different question, so we do not touch colour or sound until the story is settled.

Assembly is where every usable take gets pulled onto the timeline in script or shoot order. Nothing is judged yet. The rough cut is the first real edit: structure, order of ideas, obvious fat removed. It will feel long and unpolished, and that is correct at this point. The fine cut tightens pacing frame by frame, fixes rhythm, and settles the exact in and out points of every cut.

Picture lock is the moment the edit stops changing. It matters because everything after it, colour, sound mix, graphics, is expensive to redo. Once picture is locked we grade, mix, add titles and graphics, then export. If a client asks for a structural change after lock, the finishing work has to be redone, which is why we push for honest feedback at fine cut.

• Assembly: all usable footage laid out, nothing judged yet
• Rough cut: structure and order of ideas, deliberately loose
• Fine cut: pacing, rhythm and exact cut points
• Picture lock: the edit is final, finishing can start
• Finishing: colour, sound mix, titles, graphics
• Delivery: platform exports, captions, archive files

Colour correction and colour grading are not the same job

Colour correction fixes problems. Colour grading creates a look. They happen in that order, and confusing them is the most common reason video looks amateur despite good footage.

Correction is technical. It balances exposure so shots are not too dark or blown out, sets white balance so white objects read as white rather than orange or blue, and matches shots so a cut between two cameras does not jump. If a talking head was filmed over two hours as the daylight changed, correction is what makes the finished piece look like it was shot in one go. This stage is measurable. Scopes show whether skin tones sit in the right range and whether blacks are crushed.

Grading is creative and comes second. Warm and slightly lifted for lifestyle, cooler and higher contrast for tech, flatter and neutral for corporate or medical. A grade should support the message, not announce itself. If viewers notice the colour before the content, the grade is too heavy. Footage shot in a flat or log profile gives far more room to grade, because the camera kept more detail in highlights and shadows. Footage shot with a look baked in gives much less.

How sound is mixed and why loudness targets differ per platform

Sound is half the perceived quality of a video, and it is where cheap edits give themselves away. A mix balances three layers: dialogue, music and effects. Dialogue sits on top and stays intelligible at all times. Music supports it and ducks under speech. Effects and room tone fill gaps so silence does not feel like a dropout.

The technical part is loudness. Platforms measure loudness in LUFS, an average across the whole piece, and they normalise anything that arrives louder than their target. Making a video artificially loud does not make it play louder. It only means the platform turns it down and the dynamic range gets squashed for nothing. Different platforms sit at different targets, broadly around minus 14 LUFS for most streaming and social video, with broadcast standards sitting lower again in the UK and Europe.

Practically this means we mix once to a sensible target, keep true peaks below zero so nothing distorts on playback, and check the result on phone speakers as well as headphones. Most viewers will hear it through a phone. A mix that only works on studio monitors is a mix that fails in the real world.

• Dialogue cleaned first: noise reduction, de-essing, level matching
• Music ducked automatically under speech, not left to fight it
• Room tone laid under gaps so cuts do not sound hollow
• Loudness set to platform target, true peaks kept under zero
• Checked on phone speakers, laptop and headphones before delivery

Burned-in captions or subtitle files, and when to use each

Burned-in captions are painted into the picture and cannot be turned off. Subtitle files are separate text files the player overlays, and viewers can switch them on or off or change language. Most projects want both, for different reasons.

Burned-in wins on social feeds. Short-form video autoplays on mute, platform caption rendering is inconsistent, and a styled burned-in caption gives you full control of font, size, position and emphasis. The cost is that the text is permanent, so a typo means a re-export, and one burn cannot serve multiple languages.

Subtitle files win on long-form and anywhere accessibility or search matter. SRT is the simple universal format: timecodes and text, nothing else. VTT is the web equivalent and supports basic positioning and styling. Both are readable by search engines and by anyone using a screen reader or translation, which a burned-in caption is not. For YouTube in particular, uploading a proper subtitle file gives the platform a clean transcript rather than leaving it to automatic speech recognition, which mangles product names and industry terms.

• Burned-in: full styling control, best for Reels, Shorts and TikTok
• SRT: plain timed text, accepted almost everywhere
• VTT: web standard, supports positioning and light styling
• Upload a real subtitle file to YouTube rather than trusting auto-captions
• Keep captions to one or two lines and around 40 characters per line

What footage quality does to the edit and the price

The edit can fix a lot, but not everything, and the difference shows up in both quality and cost. Some problems are cheap to solve in post. Others are effectively unfixable and force a compromise in the cut.

Fixable: uneven exposure, wrong white balance, mild camera shake, background hum, inconsistent levels between speakers, dead air, and shots that run too long. Hard or impossible: heavy clipping where highlights are pure white with no detail, out-of-focus footage, badly clipped audio that distorted at the source, and severe wind noise on an unprotected microphone. Nothing recovers detail that was never recorded.

The single largest quality gain for most people is audio, not camera. A modest camera with a clean lapel or shotgun microphone beats an expensive camera with on-board sound every time. The second largest is shooting a little wider than the final frame, which leaves room to reframe for vertical crops without losing heads.

• Record audio on a dedicated microphone, not the camera body
• Shoot in a flat or log profile if the camera offers it, for grading headroom
• Leave headroom in the frame so vertical reframes are possible
• Avoid heavy in-camera sharpening and baked-in looks
• Keep original files, not the phone or app export

Delivery specs, file handover and version control

A finished edit is not one file. It is a set of exports matched to where the video will live, plus the source assets you should keep. Getting this right avoids re-encoding later, which always loses quality.

For almost every web and social destination the practical answer is H.264 or H.265 in an MP4 container, with AAC audio. Frame rate matches the source, usually 24, 25 or 30, and it should not be changed at export. Resolution is 1920 by 1080 or 3840 by 2160 for horizontal, and 1080 by 1920 for vertical. Platforms re-encode everything you upload, so it is better to send a high-bitrate file and let them compress it once than to compress hard yourself and have them compress a second time.

On handover we give you the platform exports, a clean master without captions or graphics burned in, the subtitle files, and the project structure if you have asked for it. Versions are numbered, never overwritten, and review happens on a timestamped review link so feedback is anchored to the exact frame. That single habit removes most of the confusion in revision rounds, because "the bit near the end" becomes "at 2:14".

• MP4 with H.264 or H.265 video and AAC audio covers nearly every destination
• Match source frame rate, never convert at export
• Deliver a clean master alongside the captioned social cuts
• Number versions, keep the old ones, never overwrite
• Give feedback with timecodes so revisions are unambiguous