Sound for film: location recording and sound design

In short

Film sound is built in two halves: what is captured on set (location or sync sound — dialogue, wild lines, room tone) and what is created afterwards (dialogue edit, ADR/dubbing, Foley, ambiences, hard effects, designed elements, music) and then balanced in a mix. The craft is half physics — sound pressure level, wavelength, distance law, phase, reverberation, sampling — and half dramaturgy: what the audience should hear, when, and how loud. Indian practice shifted decisively from studio dubbing to sync sound after *Lagaan* (2001), and Dolby 5.1/7.1/Atmos mixes are now routine for mainstream Hindi, Telugu, Tamil, Kannada and Malayalam releases. It is examined in FTII, SRFTI, mass-communication and B.Sc. Film & TV papers (sound theory, recording practice, post-production) and tested in practical rounds by production houses, OTT post facilities, news and ad studios, and audio-post houses in Mumbai, Hyderabad, Chennai, Kochi and Noida.

Explained

Sound fundamentals a sound recordist must know

• Sound is a longitudinal pressure wave; audible range roughly 20 Hz–20 kHz; speed in air ≈ 343 m/s at 20 °C (rises about 0.6 m/s per °C).
• Wavelength λ = c/f: 100 Hz is 3.43 m long, 10 kHz is 34 mm. Low frequencies bend round obstacles and leak through walls; high frequencies are directional, absorbed by air and blocked by a hand or a blanket. This single fact explains why a shotgun mic cannot be directional at bass frequencies and why traffic rumble is impossible to blanket out.
• Sound pressure level (SPL) in dB is referenced to 20 µPa. Doubling distance from a point source loses 6 dB; doubling the number of equal, uncorrelated sources adds 3 dB.
• Weighting: A-weighting (dB(A)) approximates human hearing and is used for noise limits; C-weighting is used for cinema alignment and low-frequency levels; Leq is the energy-average level over a period.
• Masking: a loud sound hides a quieter one of nearby frequency — why dialogue is carved out of music by EQ (presence 1–4 kHz) rather than by raising it endlessly.
• Haas (precedence) effect: a reflection arriving within ~30 ms is fused with the direct sound; beyond ~50 ms it is heard as a distinct echo.
• Proximity effect: directional mics boost bass as the source gets closer; corrected with a high-pass filter.
• Reverberation (RT60): time for sound to fall 60 dB after the source stops. On location it is the enemy — a reverberant take cannot be "de-reverbed" convincingly, while reverb can always be added later.

Microphones and the location kit

• Transducer types: *dynamic* (moving coil, rugged, no power, lower sensitivity — stage/field reporting); *condenser* (capacitor, needs 48 V phantom power or a battery, high sensitivity and HF detail — all film booms and lavs); *electret condenser* (lavaliers, most radio mics); *ribbon* (studio, fragile, rarely on location).
• Polar patterns: omnidirectional, cardioid, supercardioid, hypercardioid, figure-of-eight, and the shotgun (interference-tube line mic).
• Boom mics: hypercardioid/short shotgun indoors (a long shotgun in a small room picks up tube-induced colouration and off-axis reflections), long shotgun outdoors. Mounted on a boompole with a shockmount, a blimp/zeppelin windshield and a windjammer ("dead cat") for wind; aim at the mouth from above, just outside frame, following the speaking actor's lines.
• Lavalier/personal mics: 3–8 mm capsules taped or clipped on the chest, concealed under cloth with moleskin/hush-tape; problems are clothing rustle, chest-cavity boost, and loss of perspective (a lav always sounds "close").
• Radio (wireless) mics: UHF transmitter + receiver; need line of sight, clean RF scanning and antenna placement. In India, wireless microphone operation is regulated by the WPC wing of the Department of Telecommunications — some low-power bands are exempt from individual licensing by notification, others need clearance, and the exempt bands have been changed by successive notifications, so the current notification must be checked before a shoot.
• Levels and connectors: mic level ≈ −60 to −40 dBu, line level +4 dBu (0 dBu = 0.775 V rms); balanced XLR cable with differential (common-mode) rejection; AES42 for digital mics. Plug-in-power (3–5 V) on consumer jacks is not phantom power.
• Recorder/mixer: multitrack field recorder (Sound Devices, Zaxcom, Zoom, Tascam) with mic pre-amps, high-pass filters, limiters, mix-track plus ISO tracks (one per mic), timecode generator, and 32-bit-float or dual-gain recording that removes most clipping risk.
• Accessories: headphones (closed-back), slate/clapperboard, timecode sync boxes, sound blankets, carpet, apple boxes, spare batteries, rain covers, a sound cart or bag harness.

Location recording workflow, step by step

1. Script breakdown and recce: identify sync-sound scenes, dialogue-heavy interiors, and locations to avoid (flight paths, markets, construction, temple/mosque loudspeakers, generators).
2. Noise control on set: switch off fans, ACs, refrigerators, tube-light chokes; move the generator far away and behind mass; lay carpet for footsteps; blanket hard parallel surfaces; negotiate quiet windows with neighbours.
3. Mic plan per shot: boom as the primary (it carries perspective and matches the frame), lavs as safety and for wide shots, plant mics for staged positions.
4. Gain staging: set pre-amp gain so dialogue peaks around −12 to −6 dBFS, leaving headroom for a shout; engage a limiter as a safety net, not as a level control.
5. Slate and sync: clapperboard head slate (or tail slate at the end), jam-sync the camera and recorder to the same timecode at the start of the day and re-jam every few hours; note MOS (shot without sound).
6. Record the extras: room tone (30–60 s of the location with everyone silent — essential for filling dialogue edits), wild lines (off-camera re-takes of difficult lines), crowd/walla, specific props and vehicles, and location ambiences for the editor.
7. Monitoring and notes: listen on headphones throughout; keep a sound report — scene/slate/take, track layout, circled takes, problems (plane, rustle, overlap).
8. File management: Broadcast WAV (BWF) with iXML metadata (scene, take, track names, timecode), 48 kHz / 24-bit as standard, poly or mono files; two card copies plus one drive copy before the card is wiped.

Sound design and the post-production chain

• Spotting session: director, editor, sound designer and composer watch the locked cut and list every sound and music cue.
• Dialogue editing: choose the best take, fill gaps with room tone, remove clicks/mouth noise, smooth perspective and level, de-noise and de-reverb sparingly, split to character tracks for the mix.
• ADR/dubbing: re-recording dialogue to picture in a studio, in rhythm with the lip movement. India has a deep dubbing tradition — full-film dubs across Hindi, Tamil, Telugu, Malayalam, Kannada, Bengali and international languages — and this is where many sound graduates begin.
• Foley: performing human-scale sounds to picture on pits and props — footsteps (feet), moves (cloth), specifics (props). Foley makes bodies feel physically present and supplies an M&E version when dialogue is replaced.
• Ambiences/atmospheres: continuous beds (room, street, forest, railway station) that define place and allow dialogue edits to breathe.
• Hard effects and design elements: doors, guns, vehicles, impacts — from library or recorded; designed sounds are built by layering, pitch-shifting, time-stretching, reversing, convolution and worldizing (replaying a sound in a real space and re-recording it).
• Grammar of design: diegetic (inside the story world) vs non-diegetic (score, narration); sync vs asynchronous sound; sound bridge across a cut; acousmatic sound (heard, source unseen); off-screen sound to extend space beyond the frame; silence used as a dynamic value.
• Theory names examined: Michel Chion — three listening modes (causal, semantic, reduced), synchresis, "added value", the audiovisual contract; Walter Murch — encoded (speech) vs embodied (music) sound, the "law of two-and-a-half"; R. Murray Schafer — soundscape, keynote / sound signal / soundmark.

Mixing, formats and delivery

• Premix and stems: material is premixed and delivered as DM&E stems — Dialogue, Music, Effects (with Foley and ambience sub-stems). An M&E (music and effects, no dialogue) stem is mandatory for dubbed versions.
• Channel formats: mono; stereo (LR); 5.1 (L, C, R, Ls, Rs + LFE, the ".1" being a band-limited low-frequency channel, not a subwoofer feed as such); 7.1; object-based Dolby Atmos (beds plus objects panned in 3-D, rendered to the room's speaker layout); binaural and headphone mixes for OTT.
• Cinema alignment: a theatrical dubbing stage is calibrated so that −20 dBFS pink noise reads about 85 dB(C) SPL per screen channel (surround arrays about 3 dB lower), with the mixing fader at reference 7.0 — this is why a theatrical mix must be re-balanced for television and phones.
• Loudness for broadcast and OTT: measured with ITU-R BS.1770 algorithms in LUFS/LKFS; common targets are −23 LUFS (EBU R128) and −24 LKFS (ATSC A/85), with OTT platforms specifying their own figure (often around −24 to −27 LKFS) plus a true-peak ceiling of about −2 dBTP. Indian broadcasters and platforms set the exact number in their delivery spec, and these specs are revised, so the channel's current document governs.
• Delivery items: printmaster/DCP audio, DM&E stems, M&E, as-recorded dialogue, mix notes, and an AS-11/platform-specific mezzanine for TV/OTT.

Indian standards, law and set practice

• Noise Pollution (Regulation and Control) Rules, 2000 under the Environment (Protection) Act, 1986 set ambient Leq dB(A) limits — industrial 75 day / 70 night, commercial 65/55, residential 55/45, silence zone 50/40 (day 6 a.m.–10 p.m.). Loudspeakers and public address systems are generally prohibited between 10 p.m. and 6 a.m. (states may relax up to midnight for a limited number of days a year for cultural/festive occasions). These rules bind shoot-related noise and night schedules and are enforced by the state pollution control boards and police.
• Occupational noise exposure on sets and in studios now falls under the Occupational Safety, Health and Working Conditions Code, 2020 (in force 21 November 2025, replacing the Factories Act) and the rules/standards made under it; the long-standing limit in Indian practice is 90 dB(A) for 8 hours with hearing protection above it.
• Measurement instruments: sound level meters are specified by the IEC 61672 series (class 1/class 2), with corresponding Indian Standards; calibrate with an acoustic calibrator before a measurement.
• Copyright Act, 1957: a sound recording is a distinct work from the underlying literary and musical works; the producer is the author of a sound recording, so using any pre-existing song on a soundtrack needs a synchronisation licence from the music publisher/composer-lyricist side *and* a master licence from the recording owner — in practice through IPRS (authors/composers/publishers) and PPL/Novex (recordings). The 2012 amendment gave authors and composers non-assignable royalty rights in utilisations; section 31D provides a statutory licence for broadcasting. Public performance of a soundtrack in cinemas is separately licensed.
• Certification: the finished mix is submitted with the film under the Cinematograph Act, 1952 and the Cinematograph (Certification) Rules, 2024 to the CBFC; the Cinematograph (Amendment) Act, 2023 introduced age-based sub-categories (UA 7+, 13+, 16+) and anti-piracy provisions — abusive dialogue is routinely a mute/beep instruction in the sound department's hands.
• Commercials: post-production and dubbing services are taxable services; the standard GST rate of 18% applies to such services (slabs have been 5%/18%/40% since 22 September 2025).

Key points

• Sync sound - dialogue recorded live on set with the picture; *Lagaan* (2001) mainstreamed it in Hindi cinema, replacing blanket studio dubbing.
• Inverse square law - 6 dB loss per doubling of distance; the cheapest way to improve a recording is to move the mic closer.
• Boom first, lav as safety - the boom gives natural tone and perspective; a lavalier is always close-perspective and cloth-prone.
• Shotgun mic - an interference tube narrows pickup at high frequencies only; it is not "directional" at low frequencies and is not a zoom lens for sound.
• Phantom power - 48 V on pins 2 and 3 of a balanced XLR line, powering condenser mics; unrelated to plug-in power.
• Gain staging - peaks at −12 to −6 dBFS; the limiter is a safety net, not a level control.
• Room tone - 30–60 s of the location's own silence; without it dialogue edits are audible as holes.
• Wild lines and MOS - off-camera re-recorded lines; MOS means the shot has no sync sound at all.
• Timecode jam-sync - camera and recorder share one clock; re-jam periodically to avoid drift.
• BWF/iXML - WAV with embedded timecode, scene, take and track metadata; 48 kHz/24-bit is the standard.
• 3:1 rule - the second mic should be at least three times as far from the source as the first, giving about 9.5 dB separation and weak comb filtering.
• Comb filtering - periodic notches when the same sound reaches two mics with a delay Δt; first null at 1/(2Δt).
• RT60 (Sabine) - reverberation time; add reverb in post, never try to remove it from a bad take.
• DM&E and M&E - dialogue/music/effects stems; M&E (no dialogue) is required for dubbed versions.
• 5.1 vs Atmos - channel-based beds vs object-based panning rendered to the room.
• Cinema reference - −20 dBFS pink noise ≈ 85 dB(C) per screen channel; broadcast/OTT instead targets a LUFS figure.
• Masking - dialogue intelligibility is won by carving space in 1–4 kHz, not by raising the dialogue fader.
• Noise Rules, 2000 - residential limits 55 dB(A) day / 45 night; loudspeakers barred 10 p.m.–6 a.m.

Formulas and examples

Example 1 — boom versus camera mic (distance law and signal-to-noise). An actor's voice measures 72 dB SPL at 1 m. Location ambience is 50 dB(A).

• Boom at 0.5 m: L = 72 − 20 log₁₀(0.5/1) = 72 + 6.02 = 78.0 dB; SNR = 78.0 − 50 = 28 dB — clean.
• Camera-mounted mic at 3 m: L = 72 − 20 log₁₀(3) = 72 − 9.54 = 62.5 dB; SNR = 12.5 dB — noisy and roomy.
• Conclusion: the 0.5 m boom gives about 15.5 dB more voice over the same ambience — unachievable by any amount of gain or EQ.

Example 2 — generator noise against the Noise Rules. Ambience at the mic position is 52 dB(A); the genset adds 58 dB(A).

• Combined = 10 log₁₀(10^5.2 + 10^5.8) = 10 log₁₀(158,489 + 630,957) = 10 log₁₀(789,447) = 59.0 dB(A).
• Night limit in a residential zone is 45 dB(A) Leq, so the set exceeds it by 14 dB.
• Move the genset from 10 m to 40 m: 58 − 20 log₁₀(40/10) = 58 − 12.0 = 46 dB; combined with 52 dB → 10 log₁₀(158,489 + 39,811) = 53.0 dB(A). Still above the night limit, so acoustic housing or a mains/battery supply is needed.

Example 3 — card capacity, headroom and sync drift. Eight ISO tracks, 48 kHz, 24-bit, one hour.

• Per track: 48,000 × 3 = 144,000 bytes/s. Eight tracks: 1,152,000 bytes/s.
• One hour: 1,152,000 × 3,600 = 4,147,200,000 bytes ≈ 4.15 GB; a 64 GB card holds 64/4.1472 ≈ 15.4 hours.
• Dynamic range: 6.02 × 24 ≈ 144.5 dB theoretical (a real recorder achieves ~120 dB); 16-bit gives only 6.02 × 16 ≈ 96.3 dB, which is why 24-bit plus −12 dBFS headroom is the location standard.
• Drift: 24 fps audio placed on a 23.976 fps (1000/1001) timeline drifts 0.0999% — over 10 minutes, 600 × 0.000999 = 0.60 s ≈ 14 frames. Hence jam-sync or a correct pull-down.

Example 4 — two mics on one actor (3:1 rule and comb filtering). Lavalier at 0.15 m, boom at 0.60 m from the mouth.

• Path difference = 0.45 m → Δt = 0.45/343 = 1.31 ms.
• First null f₁ = 1/(2 × 0.001312) = 381 Hz, with further nulls at odd multiples (≈1.14 kHz, 1.90 kHz) — a hollow, phasey voice if both are mixed at equal level.
• 3:1 check: 0.60 = 4 × 0.15, so the distance ratio is 4:1 — it satisfies the 3:1 rule — and the boom is 20 log₁₀(4) = 12.0 dB lower on that voice; keeping one mic clearly dominant (or using only one at a time) avoids the comb.

Example 5 — taming a reverberant location room. Room 6 × 5 × 3 m → V = 90 m³; existing absorption A = 18 m² sabins.

• RT60 = 0.161 × 90 / 18 = 0.81 s — too live for dialogue.
• Hang 20 m² of sound blankets, α ≈ 0.7 → added absorption 14 m² sabins → A = 32.
• RT60 = 14.49 / 32 = 0.45 s — usable; the remaining room character can be matched in post with convolution reverb.

Where it is used

• Feature films: location sound recordist, boom operator, sound designer, dialogue editor, re-recording mixer — Dolby-certified mixing stages at Yash Raj and Sound & Vision (Mumbai), Annapurna and Ramanaidu (Hyderabad), AVM and Prasad (Chennai), Kochi and Kolkata facilities.
• OTT and television: series post, multi-language dubbing and M&E delivery for Netflix, Prime Video, JioHotstar, SonyLIV, Zee5; each platform enforces its own loudness and stem spec.
• News, sports and live events: ENG field recording, radio mics, PA and OB mixing — a large recruiter of diploma/degree sound graduates.
• Advertising and corporate video: 30-second ad mixes, VO booths, jingles; GST-invoiced studio services billed in rupees per shift.
• Games, AR/VR and animation: object-based and interactive sound design, Foley for animation (which has no production sound at all).
• Education and heritage: FTII Pune and SRFTI Kolkata sound departments, soundscape and oral-history documentation, and acoustics consultancy for studios and auditoria.

Interview questions

• Q: Why is a boom usually better than a lavalier for dialogue? A: It sits closer to the mouth in free air, captures natural perspective that matches the shot size, and avoids clothing rustle and chest resonance; the lav is a safety and wide-shot solution.
• Q: A shotgun mic is "more directional", so can I record dialogue from 4 m away? A: No — the interference tube narrows pickup only at higher frequencies, so distance brings in room reverberation and noise with a thin, off-axis tone; directivity does not undo the inverse square law.
• Q: What is room tone and why is it recorded? A: 30–60 seconds of the location's own ambience with the set silent; the dialogue editor uses it to fill pauses and joins so edits are inaudible.
• Q: What is phantom power and what needs it? A: 48 V DC sent over a balanced XLR line to power condenser microphones; dynamic mics do not need it and plug-in power on consumer jacks is a different, lower-voltage supply.
• Q: Why 48 kHz and 24-bit on location? A: 48 kHz is the picture-industry sample rate and satisfies Nyquist for a 20 kHz band; 24-bit gives about 144 dB of theoretical range, so you can record conservatively with headroom and still have a low noise floor.
• Q: What is jam-sync, and what happens if you skip it? A: Copying one timecode clock to camera and recorder so files align in the edit; without it, clocks drift and sound must be realigned shot by shot or by waveform/auto-sync.
• Q: Explain the 3:1 rule. A: When two mics cover one source, keep the farther one at least three times the distance of the nearer one, so it is about 9.5 dB down and the comb filtering from their time difference is negligible.
• Q: What are DM&E stems and why is an M&E needed? A: Separate dialogue, music and effects mixes; the music-and-effects-only version lets a film be dubbed into other languages without rebuilding the soundtrack — critical in India's multi-language releases.
• Q: Difference between diegetic and non-diegetic sound? A: Diegetic sound exists in the story world and the characters can hear it; non-diegetic sound, such as score or narration, exists only for the audience.
• Q: Why does a theatrical mix sound wrong on a laptop, and what do you do? A: The theatre is calibrated to 85 dB(C) reference with wide dynamic range, while a laptop has no low end and a far higher ambient noise floor; you make a separate near-field/OTT mix with reduced dynamics and a platform-specified LUFS level.
• Q: A take has a passing aircraft but the performance is the best. Options? A: Record a clean wild line or ADR it, use the lav if it is less affected, cover the gap with a designed ambience, or in a worst case keep the take and build a matching aircraft element in the mix so the noise reads as intentional.
• Q: What permissions and limits govern night shooting with a generator and playback speakers in a residential area? A: The Noise Pollution (Regulation and Control) Rules, 2000 set 45 dB(A) Leq at night and generally bar loudspeakers from 10 p.m. to 6 a.m., so you need local police/authority permission, generator distancing or housing, and often a schedule change.

Common mistakes

• Treating a shotgun mic as an audio zoom and recording dialogue from the camera position.
• Riding gain or relying on the limiter instead of setting a correct pre-amp level with headroom, then clipping a shout.
• Mixing boom and lav at equal level on the same voice and introducing comb filtering (ignoring the 3:1 rule).
• Forgetting room tone, wild lines or a tail slate on an MOS/problem take — all cheap on set and expensive later.
• Assuming noise reduction and de-reverberation plugins will rescue a reverberant or noisy take; they trade artefacts for noise.
• Quoting a cinema reference level (85 dB(C)) as a broadcast specification, or delivering a theatrical mix against an OTT LUFS target.
• Using a commercial song in a student or corporate film without both the synchronisation and master licences under the Copyright Act, 1957.

Related topics

Cinematography: camera, lenses and shot design · Film editing and post-production workflow · Acoustics and studio design · Music composition and background score for film · Media law, copyright and censorship in India · Broadcast engineering and OB operations