{"course":{"slug":"ai-video","title":"Video with AI","url":"https://addaly.com/learn/ai-video","level":"intermediate","track":"create"},"lesson":{"slug":"room-tone-and-spot-effects","title":"Why silence is the worst sound","url":"https://addaly.com/learn/ai-video/room-tone-and-spot-effects","position":45,"of":66,"minutes":8,"answer":"Generated clips arrive with no sound at all, and true digital silence reads as broken, so a continuous ambience bed under every shot is the single cheapest thing that makes generated footage read as filmed.","keyIdea":"Generated clips arrive with no sound at all, and true digital silence reads as broken, so a continuous ambience bed under every shot is the single cheapest thing that makes generated footage read as filmed."},"body":{"format":"markdown","text":"## What digital silence does to a viewer\n\nReal life has no silence. There is always something: air movement, distant traffic, the building, the refrigerator, your own ears. A recording with no sound at all is an experience nobody has ever had outside a laboratory.\n\nSo when a video's audio drops to absolute zero, the viewer's reaction is not \"how peaceful\". It is that something has broken. Even in a cut between two shots, a moment of true silence reads as a fault.\n\nGenerated video arrives with no audio whatsoever. Every clip is that fault, until you fix it.\n\n## The ambience bed\n\nThe fix is one continuous audio track running underneath the whole piece, unbroken across every cut. Not per-shot sound: one bed.\n\nThis does two things at once. It removes the silence, and because it does not change at the cut points, it glues shots together. A sequence of unrelated generated clips over one continuous room tone is perceived as one place. That is an enormous amount of cohesion for one track.\n\nWhere to get it:\n\n- **Record it yourself.** Thirty seconds of the actual room, or a street, or a field, on a phone. Free, specific, and better than anything generic.\n- **Freesound**, which is a large community library under various Creative Commons licences. Check the licence on each file; some require attribution and some do not permit commercial use.\n- **The BBC sound effects archive**, which is extensive and available under its own terms: read them, because they are not a blanket free-for-commercial-use licence.\n- **Your own phone held out of a window.** Genuinely the most used technique in low-budget production.\n\nLoop it, crossfade the loop point, and lay it under everything at a low level: quiet enough that you notice it only when it stops.\n\n## Spot effects\n\nOn top of the bed go the specific sounds: footsteps, a door, a cup set down, a car passing, cloth movement, a keyboard.\n\nThe rule that matters is **sync**. A sound that lands on the frame of the action is invisible and convincing. A sound two frames late is noticed. Zoom in on the waveform and place it against the picture, rather than dropping it approximately.\n\nYou need fewer than you think. Three or four well-placed effects in a ten-second shot is a lot. A wall of sound effects reads as a game rather than as footage.\n\n## Recording your own foley\n\nThe traditional answer and still the best one for anything specific. Foley is performed sound: somebody making the noise in sync with the picture.\n\nA phone and household objects cover most of it. Footsteps recorded on the actual surface. Cloth movement by rubbing a sleeve near the microphone. A cup on a table is a cup on a table. Rain on a window is rice on a sheet of paper, which is a hundred-year-old trick that still works.\n\nRecording it yourself takes twenty minutes, sounds specific rather than stock, and has no licence attached to it at all, which, on a commercial job, is worth more than the twenty minutes.\n\n## Perspective\n\nSound has distance. A wide shot should sound further away than a close-up: quieter, with more reflection and less high frequency. Matching audio perspective to shot size is a small adjustment (a level change and a slight high-frequency roll-off) and it is one of the things that separates work that sounds designed from work that sounds assembled.\n\nThe same principle applies across a cut. If you cut from inside a room to outside it, the ambience should change at the cut. If it does not, the cut reads as a glitch rather than as a move.\n\n## The level nobody sets\n\nA practical figure worth having. Dialogue sits around -12dB to -6dB peak. Ambience sits perhaps 20 to 30dB below it: low enough that it is felt rather than heard. Music under dialogue sits lower than beginners set it, usually around -25dB.\n\nIf you cannot hear the ambience at all on a laptop speaker, that is probably correct. Check on headphones.\n\n## The free path\n\nAudacity for everything, Freesound and your own phone for material, and whatever editor you are using for placement. Resolve's free tier includes a full audio page. Kdenlive and Shotcut handle multiple audio tracks. CapCut on a phone will layer audio and is enough for short-form.\n\nNothing on this page costs money, and it is the largest perceived-quality improvement available anywhere in the pipeline.\n\n---\n\nToday: record thirty seconds of the room you are in, loop it under three generated clips, and watch them with and without it."},"check":{"question":"A sequence of six unrelated generated clips is cut together. Adding one continuous ambience track underneath makes the sequence feel like a single place. Why does one track achieve that?","options":[{"text":"The ambience does not change at the cut points, so the audience infers that the shots share an acoustic space even though the pictures do not match.","correct":true},{"text":"Continuous audio masks the visual discontinuities at each cut by occupying attention that would otherwise notice them.","correct":false,"tempting":"Attributes the effect to distraction rather than to inference; viewers still see the cuts and read them as edits within one space."},{"text":"The track provides a consistent noise floor, which normalises the perceived audio quality of clips generated at different times.","correct":false,"tempting":"Describes a real levelling benefit that would apply equally to six separate ambience clips, which do not produce the same cohesion."},{"text":"Removing digital silence eliminates the impression that the file is faulty, and once that impression is gone viewers judge the sequence on its picture continuity alone.","correct":false,"tempting":"Correct about silence reading as a fault, and it explains why silence is bad rather than why one unbroken bed unifies the shots."}]},"prev":{"slug":"recording-a-real-voice","title":"A phone, a quiet room, and four minutes in Audacity","url":"https://addaly.com/learn/ai-video/recording-a-real-voice"},"next":{"slug":"music-and-its-licence","title":"Music you could defend if somebody asked","url":"https://addaly.com/learn/ai-video/music-and-its-licence"},"authorship":"Founder and series editor: Dr Mustafa Mun. Written by AI models to a written standard. Exam keys checked blind by a second model: 3,665 of 3,666 agreed. Academic review invited.","review":{"status":"Draft, open for academic review"},"cite":"Addaly (2026) 'Why silence is the worst sound', in Video with AI. Edited by M. Mun. Addaly. Available at: https://addaly.com/learn/ai-video/room-tone-and-spot-effects (Accessed: 13 October 2026).","licence":{"name":"CC BY-NC-SA 4.0","url":"https://creativecommons.org/licenses/by-nc-sa/4.0/"},"reader":{"schema":1,"title":"Why silence is the worst sound","courseTitle":"Video with AI","module":{"index":5,"slug":"sound-and-speech","title":"The half of video that is not picture"},"answer":null,"keyIdea":"Generated clips arrive with no sound at all, and true digital silence reads as broken, so a continuous ambience bed under every shot is the single cheapest thing that makes generated footage read as filmed.","hero":{"kind":"plate","image":null,"alt":null,"caption":null},"blocks":[{"t":"h2","id":"what-digital-silence-does-to-a-viewer","runs":[{"text":"What digital silence does to a viewer"}]},{"t":"p","runs":[{"text":"Real life has no silence. There is always something: air movement, distant traffic, the building, the refrigerator, your own ears. A recording with no sound at all is an experience nobody has ever had outside a laboratory."}]},{"t":"p","runs":[{"text":"So when a video’s audio drops to absolute zero, the viewer’s reaction is not “how peaceful”. It is that something has broken. Even in a cut between two shots, a moment of true silence reads as a fault."}]},{"t":"p","runs":[{"text":"Generated video arrives with no audio whatsoever. Every clip is that fault, until you fix it."}]},{"t":"h2","id":"the-ambience-bed","runs":[{"text":"The ambience bed"}]},{"t":"p","runs":[{"text":"The fix is one continuous audio track running underneath the whole piece, unbroken across every cut. Not per-shot sound: one bed."}]},{"t":"p","runs":[{"text":"This does two things at once. It removes the silence, and because it does not change at the cut points, it glues shots together. A sequence of unrelated generated clips over one continuous room tone is perceived as one place. That is an enormous amount of cohesion for one track."}]},{"t":"p","runs":[{"text":"Where to get it:"}]},{"t":"ul","items":[[{"text":"Record it yourself.","b":true},{"text":" Thirty seconds of the actual room, or a street, or a field, on a phone. Free, specific, and better than anything generic."}],[{"text":"Freesound","b":true},{"text":", which is a large community library under various Creative Commons licences. Check the licence on each file; some require attribution and some do not permit commercial use."}],[{"text":"The BBC sound effects archive","b":true},{"text":", which is extensive and available under its own terms: read them, because they are not a blanket free-for-commercial-use licence."}],[{"text":"Your own phone held out of a window.","b":true},{"text":" Genuinely the most used technique in low-budget production."}]]},{"t":"p","runs":[{"text":"Loop it, crossfade the loop point, and lay it under everything at a low level: quiet enough that you notice it only when it stops."}]},{"t":"h2","id":"spot-effects","runs":[{"text":"Spot effects"}]},{"t":"p","runs":[{"text":"On top of the bed go the specific sounds: footsteps, a door, a cup set down, a car passing, cloth movement, a keyboard."}]},{"t":"p","runs":[{"text":"The rule that matters is "},{"text":"sync","b":true},{"text":". A sound that lands on the frame of the action is invisible and convincing. A sound two frames late is noticed. Zoom in on the waveform and place it against the picture, rather than dropping it approximately."}]},{"t":"p","runs":[{"text":"You need fewer than you think. Three or four well-placed effects in a ten-second shot is a lot. A wall of sound effects reads as a game rather than as footage."}]},{"t":"h2","id":"recording-your-own-foley","runs":[{"text":"Recording your own foley"}]},{"t":"p","runs":[{"text":"The traditional answer and still the best one for anything specific. Foley is performed sound: somebody making the noise in sync with the picture."}]},{"t":"p","runs":[{"text":"A phone and household objects cover most of it. Footsteps recorded on the actual surface. Cloth movement by rubbing a sleeve near the microphone. A cup on a table is a cup on a table. Rain on a window is rice on a sheet of paper, which is a hundred-year-old trick that still works."}]},{"t":"p","runs":[{"text":"Recording it yourself takes twenty minutes, sounds specific rather than stock, and has no licence attached to it at all, which, on a commercial job, is worth more than the twenty minutes."}]},{"t":"h2","id":"perspective","runs":[{"text":"Perspective"}]},{"t":"p","runs":[{"text":"Sound has distance. A wide shot should sound further away than a close-up: quieter, with more reflection and less high frequency. Matching audio perspective to shot size is a small adjustment (a level change and a slight high-frequency roll-off) and it is one of the things that separates work that sounds designed from work that sounds assembled."}]},{"t":"p","runs":[{"text":"The same principle applies across a cut. If you cut from inside a room to outside it, the ambience should change at the cut. If it does not, the cut reads as a glitch rather than as a move."}]},{"t":"h2","id":"the-level-nobody-sets","runs":[{"text":"The level nobody sets"}]},{"t":"p","runs":[{"text":"A practical figure worth having. Dialogue sits around -12dB to -6dB peak. Ambience sits perhaps 20 to 30dB below it: low enough that it is felt rather than heard. Music under dialogue sits lower than beginners set it, usually around -25dB."}]},{"t":"p","runs":[{"text":"If you cannot hear the ambience at all on a laptop speaker, that is probably correct. Check on headphones."}]},{"t":"h2","id":"the-free-path","runs":[{"text":"The free path"}],"role":"free"},{"t":"p","runs":[{"text":"Audacity for everything, Freesound and your own phone for material, and whatever editor you are using for placement. Resolve’s free tier includes a full audio page. Kdenlive and Shotcut handle multiple audio tracks. CapCut on a phone will layer audio and is enough for short-form."}]},{"t":"p","runs":[{"text":"Nothing on this page costs money, and it is the largest perceived-quality improvement available anywhere in the pipeline."}]},{"t":"rule"},{"t":"p","runs":[{"text":"Today: record thirty seconds of the room you are in, loop it under three generated clips, and watch them with and without it."}]}],"check":{"question":[{"text":"A sequence of six unrelated generated clips is cut together. Adding one continuous ambience track underneath makes the sequence feel like a single place. Why does one track achieve that?"}],"options":[{"text":[{"text":"The ambience does not change at the cut points, so the audience infers that the shots share an acoustic space even though the pictures do not match."}],"correct":true,"tempting":null},{"text":[{"text":"Continuous audio masks the visual discontinuities at each cut by occupying attention that would otherwise notice them."}],"correct":false,"tempting":[{"text":"Attributes the effect to distraction rather than to inference; viewers still see the cuts and read them as edits within one space."}]},{"text":[{"text":"The track provides a consistent noise floor, which normalises the perceived audio quality of clips generated at different times."}],"correct":false,"tempting":[{"text":"Describes a real levelling benefit that would apply equally to six separate ambience clips, which do not produce the same cohesion."}]},{"text":[{"text":"Removing digital silence eliminates the impression that the file is faulty, and once that impression is gone viewers judge the sequence on its picture continuity alone."}],"correct":false,"tempting":[{"text":"Correct about silence reading as a fault, and it explains why silence is bad rather than why one unbroken bed unifies the shots."}]}],"order":[1,0,2,3],"why":null},"prev":{"slug":"recording-a-real-voice","title":"A phone, a quiet room, and four minutes in Audacity","minutes":9},"next":{"slug":"music-and-its-licence","title":"Music you could defend if somebody asked","minutes":9}}}