Listen before choosing the picture
Walk the proposed interview area while the facility is operating normally. Listen at head height, not from the doorway. Identify continuous noise, cycles, tones, impacts, moving vehicles, public-address messages and alarms. Ask the site contact what may start during the interview and whether any equipment can be paused through an approved operational decision.
Then record a short test and monitor it on isolating headphones. Human hearing adapts to steady noise; a microphone records it without that mental filtering. Check the test for intelligibility, distortion, radio interference and low-frequency vibration. A level meter on screen can look healthy while the voice is acoustically buried.
Move the interview, even if only a few metres
Distance and barriers can change the result more than changing equipment. Look for a solid wall between the speaker and the source, a doorway that can close, a corner outside the direct airflow, or a position where the loudest source is behind the microphone rather than behind the contributor. Avoid small metal rooms that add reflections to the noise.
Keep enough visual context to establish the facility. That might mean filming a clean interview in an adjacent control area and collecting the louder operational pictures separately. The final edit can remain honest: establish the location, show the process, then hear the specialist clearly. There is no rule that every sentence must be recorded beside the machine it describes.
Put the microphone close to the voice
Speech-to-noise ratio is the core problem. Shure’s technical guidance makes the relationship plain: move the microphone closer to the talker, reduce the room noise, or do both. A highly directional microphone placed too far away still receives the machinery and reflections along with the voice.
For an interview, place the chosen microphone as close as the frame, clothing, PPE and safety controls reasonably allow. Aim its most sensitive direction towards the mouth and its least sensitive direction towards the dominant noise when the microphone design allows. Fit suitable wind, breath or contact-noise protection for the position. Check that a clip does not rub against a high-visibility jacket, zip, lanyard or respirator strap.
We used close positioning and appropriate protection/filtering at Coryton because the operational background was exceptionally loud. The exact microphone is less useful than the principle: capture more direct voice at source. Equipment choice should follow the contributor, clothing, radio environment, noise spectrum and permitted working position.
Record a primary track and a genuinely independent backup
A backup helps only when it can survive a different failure. Two tracks from the same poorly placed microphone do not protect against clothing rustle or acoustic masking. Depending on the production, independence might mean a second close microphone, local recording at the transmitter, a separate recorder, or a second system monitored by another crew member.
Before the answer begins:
- confirm both tracks are receiving the intended microphone;
- listen for radio hits, electrical interference and cable movement;
- leave headroom for sudden emphasis;
- check the contributor can turn naturally without moving off-mic;
- record a spoken identification for the person and location.
Monitor throughout. A clean rehearsal does not prove that a nearby drive, fan or pump will stay in the same state.
Structure the interview for difficult sound
Ask short questions and encourage complete answers. Long, overlapping conversation is harder to edit when every cut exposes a changing machine cycle. Leave a beat before the contributor answers and another after they finish. If an impact, alarm or vehicle masks a key phrase, repeat the answer immediately while the thought is fresh.
Capture names, technical terms and figures a second time. Ask the client representative to verify spelling and meaning. Industrial films often contain terms an editor cannot infer reliably from context.
If the contributor must wear hearing protection, agree how questions will be delivered without defeating the site’s controls. Never ask someone to remove required protection for the recording. HSE guidance says hearing protection must be worn in designated zones and warns that overprotection can also make communication and warning signals difficult; the site’s competent person should determine the safe arrangement.
Give the editor clean material to work with
Record at least 30 seconds of the background with nobody speaking, ideally for each operating state heard during the interview. Capture the relevant machinery separately for picture and natural sound. Make notes when a fan starts, a vehicle passes or an answer is repeated.
Dialogue cleanup may combine tonal hum removal, broadband noise reduction, spectral repair, equalisation and careful level control. Apply it in stages and compare constantly with the unprocessed voice. Aggressive processing can create watery consonants, pumping ambience or a voice that no longer matches the room.
The edit has three honest options when a sentence cannot be recovered:
- use a cleaner repeated take;
- replace the section with verified on-screen text or a newly recorded explanation;
- remove it.
Subtitles improve accessibility and help with unfamiliar terminology, but they are not permission to publish needlessly unintelligible dialogue. Viewers still use tone, emphasis and the sound of the place to judge the speaker.
A 15-minute on-site sound test
| Minute | Check | Decision |
|---|---|---|
| 0–3 | Listen through one full operating cycle | Continuous, cyclic or intermittent noise? |
| 3–6 | Test two safe positions | Which gives the best voice-to-noise balance? |
| 6–9 | Fit and monitor the close microphone | Any PPE, clothing, wind or radio issue? |
| 9–12 | Record a sample answer at real speaking volume | Is every word intelligible without cleanup? |
| 12–15 | Trigger or wait for likely interruptions | Pause, relocate or plan repeated takes? |
Do this before lighting is final. Moving a complete interview setup after discovering a fan cycle wastes more time than an early sound test.
