48-hour brand film: the sentence, the hiss and the 30-second licence
Cut Republic must deliver a three-minute brand film for a handloom co-operative in 48 hours. The footage came from two shooters: Camera A at 1080p, 25 fps, with the main interview and sync sound; Camera B at 4K, 30 fps, handheld cutaways of looms and yarn, no usable sync sound. The co-operative's lawyer has asked for one sentence to be removed from the founder's answer. The sentence sits in the middle of a single continuous take, and the founder's hands are in frame throughout. The voiceover was recorded on a laptop microphone in a room with a running air conditioner. The music bed is licensed for 30 seconds only, and the cut runs 2 minutes 52 seconds. Delivery is required both as a file for a 4K projector at a screening and as a YouTube upload.
1. Lay out the order of operations across the 48 hours, with rough hour blocks, and justify why picture decisions must come before audio and graphics work.
Read the worked answer
Hours 0–2 - safety and setup. Two backups on separate drives, verify the cards, build the project at 25 fps, transcode Camera B's 30 fps to the timeline, and write down the exact timecode in and out of the sentence to be removed.
Hours 2–6 - rough cut and the legal cut. Remove the sentence, assemble, temp music, running time locked at 2:52. Decide the music problem now, because it changes the edit rhythm.
Hours 6–10 - client review. One watermarked review link, one round of notes, a stated deadline of 4 hours. Get sign-off in writing.
Hours 10–16 - picture lock. Revisions, then freeze. No further picture changes for any reason.
Hours 16–26 - conform, titles, graphics, grade. Everything that depends on where cuts fall.
Hours 26–36 - audio. Voiceover cleanup, music placement, mix, loudness to about −14 LUFS for YouTube.
Hours 36–44 - QC and export. Watch on a TV and a phone, check captions, export 4K master plus 1080p H.264, name files properly.
Hours 44–48 - buffer. This block is not padding; it is where the client's surprise request lands.
The justification for the ordering is simple economics: any minute spent on audio, tracking or grading before picture lock is a minute spent twice. The sentence removal is the one task that can move a cut, so it must happen before anything else is built on top of the cut.
2. How do you handle the two frame rates, and what does that decision let you do about the removed sentence?
Read the worked answer
The timeline is 25 fps because Camera A carries all sync dialogue and India delivers at 25.
Camera B's 30 fps clips, dropped unmodified into a 25 fps timeline, will stutter - frames are duplicated and dropped unevenly. There are only two honest treatments:
- Interpret at 25 fps (a true 20% slow-motion). The loom and yarn shots become graceful slow motion, which suits handloom material perfectly, and there is no stutter. Cost: everything on B runs slower than it was shot, so it can never be used for anything that must match real time.
- Retime to real speed, accepting a small quality loss, if a B shot must feel live.
Rule: B is B-roll only. It has no usable sync sound, so it is never used for a talking face or an action that must match a line of dialogue.
The removed sentence changes the calculation: Camera A is only 1080p, so a digital push-in to hide a jump cut will soften visibly at a screening. Camera B's 4K, however, gives you room to punch in on hands at the loom and hold a two-second cutaway without any resolution loss. The 30 fps slow motion makes that cutaway look deliberate rather than like a patch. So the frame-rate decision is what makes the cover shot work.
3. Describe how you would remove the sentence, clean the voiceover, and solve the 30-second music licence - including what you would refuse to do.
Read the worked answer
The sentence. There is no clean way to delete sound from the middle of a continuous take and keep the picture still. Use the least visible method available, in this priority:
- Cut on a natural gesture or blink, then cover 2–3 seconds with the 4K loom-hands cutaway in slow motion - hands are already in frame, so the audience reads it as a reaction shot.
- J-cut the audio so the founder's next word arrives before the cutaway ends.
- If a jump remains, reframe slightly across the cut only if the shot size tolerates it. Refuse: a dissolve, a freeze frame, or a musical sting to hide the join - all three announce that something was removed, which is the opposite of what the lawyer wants.
The voiceover. Chain: noise print from the AC hum (it is broadband, not just 50 Hz), subtractive reduction of no more than 6–10 dB, notch at 50/100/150 Hz, high-pass at 90 Hz, de-ess, light compression, limiter. But be honest about the ceiling - a laptop mic in a room with a running AC has a noise floor that no plugin will rescue. With 48 hours available, re-record it: 20 minutes in a quiet room on a decent phone, or 30 minutes in a local studio, is cheaper than an hour of rescue and it will sound better. Re-record if more than about 30% of the VO is affected.
The music. A 30-second licence is a 30-second licence. Legitimate options: use the licensed track as an intro and outro bookend and carry the middle on loom sound and the founder's voice; call the licensor for an extension (usually cheap for a co-operative); or replace it with a fully cleared library cue. Refuse: looping, pitch-shifting or time-stretching the 30 seconds to cover three minutes - that breaches the licence and YouTube's Content ID will find it and monetise or mute the video. File a cue sheet either way.