{"course":{"slug":"ai-video","title":"Video with AI","url":"https://addaly.com/learn/ai-video","level":"intermediate","track":"create"},"lesson":{"slug":"matching-colour-across-generations","title":"Thirty renders, one world","url":"https://addaly.com/learn/ai-video/matching-colour-across-generations","position":53,"of":66,"minutes":10,"answer":"Each generation makes its own independent decisions about colour and contrast, so a piece is unified by a two-stage process (shot matching to bring every clip to a neutral common ground.","keyIdea":"Each generation makes its own independent decisions about colour and contrast, so a piece is unified by a two-stage process (shot matching to bring every clip to a neutral common ground, then one grade over all of it) and never by grading each clip to taste."},"body":{"format":"markdown","text":"## Why every clip is a different colour\n\nNothing carries colour decisions between generations. Two clips from the same still, the same prompt and the same model will differ slightly in white balance, contrast and saturation, because those properties emerge from the sample rather than being set by a parameter.\n\nAcross thirty clips generated over a week, that variation is large. On a timeline it reads as amateurism more reliably than almost any other fault, because audiences associate consistent colour with produced work.\n\n## Two stages, in this order\n\nThe mistake is to grade each shot until it looks nice. That produces thirty attractive shots that do not belong together.\n\n**Stage one: match.** Bring every clip to a neutral common ground. Same black level, same white point, same rough contrast. Nobody sees this stage; its output looks flat and boring and correct.\n\n**Stage two: grade.** One look, applied across everything at once, on an adjustment layer. This is where the piece gets its colour.\n\nThe reason for the order: a look applied to inconsistent clips amplifies the inconsistency, because a contrast curve pushes two different starting points further apart.\n\n> Figure: What sits above the clips on a finished timeline. Layers, top to bottom: Grain, one pass (A single layer over the whole sequence, set so you can only see it on a paused frame. Thirty separate grains is worse than none, because now the cuts have one more thing changing across them.); The grade, one layer (The look from your look file, on one adjustment layer across everything. Modest: a strong grade on generated footage draws attention to the footage.); Shot matching, per clip (Black level first, then white point, then mid-tones, then saturation, read off the waveform rather than off your eyes. Nobody sees this stage; its output is flat and correct.); The generations (Trimmed of the settle and the drift, conformed to one project frame rate, and otherwise untouched.). The order is the lesson. A look applied to clips that have not been matched amplifies the mismatch, because a contrast curve pushes two different starting points further apart. Grading each shot until it looks nice gives you thirty attractive shots that do not belong together.\n\n## Doing stage one\n\nPick a hero shot (the one closest to what you want, usually a well-exposed mid-shot) and match everything to it.\n\nIn DaVinci Resolve's free tier:\n\n1. Put the hero shot and the shot you are matching side by side, using the split-screen view on the colour page.\n2. Right-click the target clip and use **Shot Match** to that hero. It does most of the work in one click.\n3. Fix what it got wrong by hand, in this order: **black level first**, then white point, then mid-tone brightness, then saturation.\n\nBlack level first matters. If the blacks do not sit at the same level, every subsequent adjustment is fighting a mismatch you have not fixed. Use the waveform scope: find the bottom of the trace in each clip and line them up.\n\nKdenlive and Shotcut have colour-balance and curve filters but no shot-match. The manual method works: set the scopes side by side, match blacks, match whites, match saturation. Slower, same result.\n\n## Reading scopes instead of your eyes\n\nYour eyes adapt within seconds, which makes them useless for comparison. After thirty seconds of looking at a warm shot, a neutral one looks blue.\n\nScopes do not adapt:\n\n- **Waveform** shows brightness across the frame. Use it for black level and white point.\n- **Parade** shows red, green and blue separately. Use it for colour cast: if the blue trace sits higher at the bottom than the red, the shadows are blue.\n- **Vectorscope** shows hue and saturation. Skin tones fall along a specific line on it, which is the most useful single fact in colour correction: if your skin tones sit on the skin line, the shot is close to right.\n\nResolve's free tier has all three. This is professional colour instrumentation at no cost, and most people using it never open them.\n\n## Stage two, the grade\n\nOne adjustment layer over the whole timeline, carrying the look from your look file. Typically:\n\n- A contrast curve, usually lifting the blacks slightly so nothing is pure black.\n- A colour shift: warm highlights and cool shadows is the most common, and the most over-used.\n- A saturation adjustment, usually downward.\n\nKeep it modest. A strong grade on generated footage draws attention to the footage, which is the opposite of what you want.\n\n## The secondary you will eventually need\n\nWhen one element is wrong while the rest of the frame is right (a garment that came out a different blue in one shot) a global correction cannot help.\n\nA secondary isolates a colour range or a shape and corrects only that. Resolve's free tier supports qualifiers and windows, including tracking a window across a moving subject.\n\nIt is fiddly, it is imprecise on anything with motion blur or soft edges, and it is sometimes the only thing between you and regenerating the shot. Budget an hour when you need one, and treat needing one as a signal that your still-checking pass failed.\n\n## Check on more than one screen\n\nA laptop screen, a phone and, if possible, a television. Generated footage graded on a bright laptop routinely looks crushed and muddy on a phone in daylight, which is where most of your audience is.\n\nIf you can only check on one screen, check on the phone.\n\n---\n\nToday: open the waveform scope, put three of your clips on a timeline, and line up their black levels. Do nothing else and watch how much more like one piece they look."},"check":{"question":"An editor grades each generated clip individually until each looks attractive, then plays the sequence. It reads as inconsistent. What was the process error?","options":[{"text":"The clips should have been graded in sequence order rather than individually, so each could be matched to the one before it as the work progressed.","correct":false,"tempting":"Chained matching accumulates drift across thirty clips and still lacks a common reference to return to."},{"text":"Individual grades were applied to the clips rather than to an adjustment layer, so they cannot be revised together as a group later.","correct":false,"tempting":"Names a real workflow inconvenience that affects revising the work rather than why the result looks inconsistent."},{"text":"Visual adaptation means the editor's eyes adjusted between clips, so each grade was made relative to the previous shot rather than to an absolute reference, compounding the differences across the timeline.","correct":false,"tempting":"Describes a genuine reason eyes are unreliable, and the inconsistency would remain even with perfect vision because no common ground was established."},{"text":"Grading to taste optimises each shot separately, so nothing brings the clips to a common ground before a shared look is applied on top.","correct":true}]},"prev":{"slug":"where-a-cut-hides","title":"Cut on motion, and one frame before the failure","url":"https://addaly.com/learn/ai-video/where-a-cut-hides"},"next":{"slug":"grain-texture-and-defocus","title":"Putting back the texture the encoder removed","url":"https://addaly.com/learn/ai-video/grain-texture-and-defocus"},"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) 'Thirty renders, one world', in Video with AI. Edited by M. Mun. Addaly. Available at: https://addaly.com/learn/ai-video/matching-colour-across-generations (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":"Thirty renders, one world","courseTitle":"Video with AI","module":{"index":6,"slug":"the-edit","title":"From generations to a finished piece"},"answer":null,"keyIdea":"Each generation makes its own independent decisions about colour and contrast, so a piece is unified by a two-stage process (shot matching to bring every clip to a neutral common ground, then one grade over all of it) and never by grading each clip to taste.","hero":{"kind":"figure","figure":{"t":"figure","label":"Figure","figure":{"type":"stack","title":"What sits above the clips on a finished timeline","layers":[{"label":"Grain, one pass","detail":"A single layer over the whole sequence, set so you can only see it on a paused frame. Thirty separate grains is worse than none, because now the cuts have one more thing changing across them."},{"label":"The grade, one layer","detail":"The look from your look file, on one adjustment layer across everything. Modest: a strong grade on generated footage draws attention to the footage."},{"label":"Shot matching, per clip","detail":"Black level first, then white point, then mid-tones, then saturation, read off the waveform rather than off your eyes. Nobody sees this stage; its output is flat and correct."},{"label":"The generations","detail":"Trimmed of the settle and the drift, conformed to one project frame rate, and otherwise untouched."}],"caption":"The order is the lesson. A look applied to clips that have not been matched amplifies the mismatch, because a contrast curve pushes two different starting points further apart. Grading each shot until it looks nice gives you thirty attractive shots that do not belong together."},"alt":"Figure: What sits above the clips on a finished timeline. Layers, top to bottom: Grain, one pass (A single layer over the whole sequence, set so you can only see it on a paused frame. Thirty separate grains is worse than none, because now the cuts have one more thing changing across them.); The grade, one layer (The look from your look file, on one adjustment layer across everything. Modest: a strong grade on generated footage draws attention to the footage.); Shot matching, per clip (Black level first, then white point, then mid-tones, then saturation, read off the waveform rather than off your eyes. Nobody sees this stage; its output is flat and correct.); The generations (Trimmed of the settle and the drift, conformed to one project frame rate, and otherwise untouched.). The order is the lesson. A look applied to clips that have not been matched amplifies the mismatch, because a contrast curve pushes two different starting points further apart. Grading each shot until it looks nice gives you thirty attractive shots that do not belong together."}},"blocks":[{"t":"h2","id":"why-every-clip-is-a-different-colour","runs":[{"text":"Why every clip is a different colour"}]},{"t":"p","runs":[{"text":"Nothing carries colour decisions between generations. Two clips from the same still, the same prompt and the same model will differ slightly in white balance, contrast and saturation, because those properties emerge from the sample rather than being set by a parameter."}]},{"t":"p","runs":[{"text":"Across thirty clips generated over a week, that variation is large. On a timeline it reads as amateurism more reliably than almost any other fault, because audiences associate consistent colour with produced work."}]},{"t":"h2","id":"two-stages-in-this-order","runs":[{"text":"Two stages, in this order"}]},{"t":"p","runs":[{"text":"The mistake is to grade each shot until it looks nice. That produces thirty attractive shots that do not belong together."}]},{"t":"p","runs":[{"text":"Stage one: match.","b":true},{"text":" Bring every clip to a neutral common ground. Same black level, same white point, same rough contrast. Nobody sees this stage; its output looks flat and boring and correct."}]},{"t":"p","runs":[{"text":"Stage two: grade.","b":true},{"text":" One look, applied across everything at once, on an adjustment layer. This is where the piece gets its colour."}]},{"t":"p","runs":[{"text":"The reason for the order: a look applied to inconsistent clips amplifies the inconsistency, because a contrast curve pushes two different starting points further apart."}]},{"t":"h2","id":"doing-stage-one","runs":[{"text":"Doing stage one"}]},{"t":"p","runs":[{"text":"Pick a hero shot (the one closest to what you want, usually a well-exposed mid-shot) and match everything to it."}]},{"t":"p","runs":[{"text":"In DaVinci Resolve’s free tier:"}]},{"t":"ol","start":1,"items":[[{"text":"Put the hero shot and the shot you are matching side by side, using the split-screen view on the colour page."}],[{"text":"Right-click the target clip and use "},{"text":"Shot Match","b":true},{"text":" to that hero. It does most of the work in one click."}],[{"text":"Fix what it got wrong by hand, in this order: "},{"text":"black level first","b":true},{"text":", then white point, then mid-tone brightness, then saturation."}]]},{"t":"p","runs":[{"text":"Black level first matters. If the blacks do not sit at the same level, every subsequent adjustment is fighting a mismatch you have not fixed. Use the waveform scope: find the bottom of the trace in each clip and line them up."}]},{"t":"p","runs":[{"text":"Kdenlive and Shotcut have colour-balance and curve filters but no shot-match. The manual method works: set the scopes side by side, match blacks, match whites, match saturation. Slower, same result."}]},{"t":"h2","id":"reading-scopes-instead-of-your-eyes","runs":[{"text":"Reading scopes instead of your eyes"}]},{"t":"p","runs":[{"text":"Your eyes adapt within seconds, which makes them useless for comparison. After thirty seconds of looking at a warm shot, a neutral one looks blue."}]},{"t":"p","runs":[{"text":"Scopes do not adapt:"}]},{"t":"ul","items":[[{"text":"Waveform","b":true},{"text":" shows brightness across the frame. Use it for black level and white point."}],[{"text":"Parade","b":true},{"text":" shows red, green and blue separately. Use it for colour cast: if the blue trace sits higher at the bottom than the red, the shadows are blue."}],[{"text":"Vectorscope","b":true},{"text":" shows hue and saturation. Skin tones fall along a specific line on it, which is the most useful single fact in colour correction: if your skin tones sit on the skin line, the shot is close to right."}]]},{"t":"p","runs":[{"text":"Resolve’s free tier has all three. This is professional colour instrumentation at no cost, and most people using it never open them."}]},{"t":"h2","id":"stage-two-the-grade","runs":[{"text":"Stage two, the grade"}]},{"t":"p","runs":[{"text":"One adjustment layer over the whole timeline, carrying the look from your look file. Typically:"}]},{"t":"ul","items":[[{"text":"A contrast curve, usually lifting the blacks slightly so nothing is pure black."}],[{"text":"A colour shift: warm highlights and cool shadows is the most common, and the most over-used."}],[{"text":"A saturation adjustment, usually downward."}]]},{"t":"p","runs":[{"text":"Keep it modest. A strong grade on generated footage draws attention to the footage, which is the opposite of what you want."}]},{"t":"h2","id":"the-secondary-you-will-eventually-need","runs":[{"text":"The secondary you will eventually need"}]},{"t":"p","runs":[{"text":"When one element is wrong while the rest of the frame is right (a garment that came out a different blue in one shot) a global correction cannot help."}]},{"t":"p","runs":[{"text":"A secondary isolates a colour range or a shape and corrects only that. Resolve’s free tier supports qualifiers and windows, including tracking a window across a moving subject."}]},{"t":"p","runs":[{"text":"It is fiddly, it is imprecise on anything with motion blur or soft edges, and it is sometimes the only thing between you and regenerating the shot. Budget an hour when you need one, and treat needing one as a signal that your still-checking pass failed."}]},{"t":"h2","id":"check-on-more-than-one-screen","runs":[{"text":"Check on more than one screen"}]},{"t":"p","runs":[{"text":"A laptop screen, a phone and, if possible, a television. Generated footage graded on a bright laptop routinely looks crushed and muddy on a phone in daylight, which is where most of your audience is."}]},{"t":"p","runs":[{"text":"If you can only check on one screen, check on the phone."}]},{"t":"rule"},{"t":"p","runs":[{"text":"Today: open the waveform scope, put three of your clips on a timeline, and line up their black levels. Do nothing else and watch how much more like one piece they look."}]}],"check":{"question":[{"text":"An editor grades each generated clip individually until each looks attractive, then plays the sequence. It reads as inconsistent. What was the process error?"}],"options":[{"text":[{"text":"The clips should have been graded in sequence order rather than individually, so each could be matched to the one before it as the work progressed."}],"correct":false,"tempting":[{"text":"Chained matching accumulates drift across thirty clips and still lacks a common reference to return to."}]},{"text":[{"text":"Individual grades were applied to the clips rather than to an adjustment layer, so they cannot be revised together as a group later."}],"correct":false,"tempting":[{"text":"Names a real workflow inconvenience that affects revising the work rather than why the result looks inconsistent."}]},{"text":[{"text":"Visual adaptation means the editor’s eyes adjusted between clips, so each grade was made relative to the previous shot rather than to an absolute reference, compounding the differences across the timeline."}],"correct":false,"tempting":[{"text":"Describes a genuine reason eyes are unreliable, and the inconsistency would remain even with perfect vision because no common ground was established."}]},{"text":[{"text":"Grading to taste optimises each shot separately, so nothing brings the clips to a common ground before a shared look is applied on top."}],"correct":true,"tempting":null}],"order":[2,0,3,1],"why":null},"prev":{"slug":"where-a-cut-hides","title":"Cut on motion, and one frame before the failure","minutes":8},"next":{"slug":"grain-texture-and-defocus","title":"Putting back the texture the encoder removed","minutes":8}}}