{"course":{"slug":"ai-video","title":"Video with AI","url":"https://addaly.com/learn/ai-video","level":"intermediate","track":"create"},"lesson":{"slug":"selects-and-the-assembly","title":"Choosing takes before you start cutting","url":"https://addaly.com/learn/ai-video/selects-and-the-assembly","position":51,"of":66,"minutes":8,"answer":"Selecting takes and assembling a rough cut are separate passes with different judgements.","keyIdea":"Selecting takes and assembling a rough cut are separate passes with different judgements, and collapsing them is why edits stall: you cannot decide whether a shot works in the sequence while you are still deciding whether the shot works."},"body":{"format":"markdown","text":"## Four passes, in order\n\nEditors work in passes because each one asks a different question and mixing them produces paralysis.\n\n1. **Selects.** For each shot, which take? Judge the take alone.\n2. **Assembly.** Put the chosen takes in order, full length, no trimming. Does the sequence work?\n3. **Rough cut.** Trim to length. Find the cut points.\n4. **Fine cut.** Frame-level adjustment, then colour, then sound, then titles.\n\nBeginners try to do all four at once and get stuck on shot three for an hour. The passes exist because each one's decisions are only makeable once the previous one is settled.\n\n## The selects pass\n\nYou generated four takes of shot nine. Watch all four, in full, before choosing.\n\nWhat you are judging, in this order:\n\n- **Did it do the thing the shot is for?** Not whether it is pretty. If shot nine exists to show her noticing the door, the take where she notices the door wins even if another is prettier.\n- **Where does it break, and when?** Note the timecode. A take that fails at six seconds is fine if the shot is four seconds long.\n- **Does it match its neighbours?** Light direction, wardrobe, screen direction.\n- **Then aesthetics.**\n\nMark the chosen take and keep the rejects. A reject often becomes a cutaway, an insert, or a source for a repaired frame later.\n\nWrite the choice down next to the shot on your list. Do not rely on file names alone; `s09_take4` does not tell you why.\n\n## The stringout\n\nPut every selected take end to end, untrimmed, in script order. This is a stringout and it is the most useful ugly thing in editing.\n\nIt will be far too long and it will feel wrong. That is correct. What it tells you, which nothing else does:\n\n- Whether the sequence makes sense at all.\n- Which shots are missing. You will find two or three, every time, and finding them now costs a generation rather than a re-edit.\n- Which shots are unnecessary. You will find more of these than missing ones.\n- Where the piece sags.\n\nWatch it once, all the way through, without stopping to fix anything. Take notes on paper. Stopping to fix is the beginning of the hour on shot three.\n\n## Cutting the length down\n\nFrom stringout to rough cut, the operations are almost all subtraction:\n\n- **Trim the head and tail of every clip.** Roughly a quarter-second off the front for the settle, and a second off the back for the drift. This is not a stylistic choice; it is removing the parts where the generation announces itself.\n- **Cut the shots that were doing nothing.** If removing a shot does not hurt, it was not working.\n- **Shorten everything.** Almost every rough cut is improved by being ten per cent shorter. Commercial editing runs two to four seconds a shot; your instinct for shot length is probably long.\n\n## Pace is a pattern, not a speed\n\nA sequence of shots all the same length reads as mechanical. Vary them: a long establishing shot, three short ones, a longer one to let something land.\n\nThe simplest useful pattern is to let the shot after a fast run breathe. Three quick cuts then a held shot gives the audience somewhere to arrive, and it is the difference between energy and noise.\n\nIn generative work this pattern has a practical bonus: short shots hide drift, and the one long shot can be the one you generated most carefully.\n\n## Keeping the decision record\n\nAt the end of the assembly, your shot list should carry, per shot: the chosen take, the reason, and the intended length. That document is what lets you answer a client's note about shot nine three weeks later without rewatching everything.\n\nIt is also what tells you, honestly, how many generations the project actually took: the number your next quote depends on.\n\n---\n\nToday: build a stringout of everything you have generated for one project and watch it once, straight through, taking notes on paper without touching the timeline."},"check":{"question":"What does a stringout (every selected take, untrimmed, in order) reveal that a partially trimmed rough cut does not?","options":[{"text":"The true final duration of the piece, which cannot be estimated once clips have been trimmed to different lengths.","correct":false,"tempting":"Reverses the relationship: a stringout is always far longer than the finished piece and tells you little about final duration."},{"text":"Where each generation begins to drift, since the untrimmed tails are still present and can be compared across shots.","correct":false,"tempting":"A real incidental benefit that belongs to the selects pass, where each take was already watched in full."},{"text":"How well the shots colour-match, because viewing them at full length under identical conditions makes differences in grade and exposure easier to detect than brief trimmed fragments would.","correct":false,"tempting":"Colour comparison is genuinely easier on longer clips, and it is a later pass whose decisions do not depend on structure being settled."},{"text":"Whether the sequence holds together and which shots are missing or unnecessary, before any trimming decisions have foreclosed the structure.","correct":true}]},"prev":{"slug":"conforming-and-project-setup","title":"Decide the frame rate before you import anything","url":"https://addaly.com/learn/ai-video/conforming-and-project-setup"},"next":{"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"},"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) 'Choosing takes before you start cutting', in Video with AI. Edited by M. Mun. Addaly. Available at: https://addaly.com/learn/ai-video/selects-and-the-assembly (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":"Choosing takes before you start cutting","courseTitle":"Video with AI","module":{"index":6,"slug":"the-edit","title":"From generations to a finished piece"},"answer":null,"keyIdea":"Selecting takes and assembling a rough cut are separate passes with different judgements, and collapsing them is why edits stall: you cannot decide whether a shot works in the sequence while you are still deciding whether the shot works.","hero":{"kind":"plate","image":null,"alt":null,"caption":null},"blocks":[{"t":"h2","id":"four-passes-in-order","runs":[{"text":"Four passes, in order"}]},{"t":"p","runs":[{"text":"Editors work in passes because each one asks a different question and mixing them produces paralysis."}]},{"t":"ol","start":1,"items":[[{"text":"Selects.","b":true},{"text":" For each shot, which take? Judge the take alone."}],[{"text":"Assembly.","b":true},{"text":" Put the chosen takes in order, full length, no trimming. Does the sequence work?"}],[{"text":"Rough cut.","b":true},{"text":" Trim to length. Find the cut points."}],[{"text":"Fine cut.","b":true},{"text":" Frame-level adjustment, then colour, then sound, then titles."}]]},{"t":"p","runs":[{"text":"Beginners try to do all four at once and get stuck on shot three for an hour. The passes exist because each one’s decisions are only makeable once the previous one is settled."}]},{"t":"h2","id":"the-selects-pass","runs":[{"text":"The selects pass"}]},{"t":"p","runs":[{"text":"You generated four takes of shot nine. Watch all four, in full, before choosing."}]},{"t":"p","runs":[{"text":"What you are judging, in this order:"}]},{"t":"ul","items":[[{"text":"Did it do the thing the shot is for?","b":true},{"text":" Not whether it is pretty. If shot nine exists to show her noticing the door, the take where she notices the door wins even if another is prettier."}],[{"text":"Where does it break, and when?","b":true},{"text":" Note the timecode. A take that fails at six seconds is fine if the shot is four seconds long."}],[{"text":"Does it match its neighbours?","b":true},{"text":" Light direction, wardrobe, screen direction."}],[{"text":"Then aesthetics.","b":true}]]},{"t":"p","runs":[{"text":"Mark the chosen take and keep the rejects. A reject often becomes a cutaway, an insert, or a source for a repaired frame later."}]},{"t":"p","runs":[{"text":"Write the choice down next to the shot on your list. Do not rely on file names alone; "},{"text":"s09_take4","code":true},{"text":" does not tell you why."}]},{"t":"h2","id":"the-stringout","runs":[{"text":"The stringout"}]},{"t":"p","runs":[{"text":"Put every selected take end to end, untrimmed, in script order. This is a stringout and it is the most useful ugly thing in editing."}]},{"t":"p","runs":[{"text":"It will be far too long and it will feel wrong. That is correct. What it tells you, which nothing else does:"}]},{"t":"ul","items":[[{"text":"Whether the sequence makes sense at all."}],[{"text":"Which shots are missing. You will find two or three, every time, and finding them now costs a generation rather than a re-edit."}],[{"text":"Which shots are unnecessary. You will find more of these than missing ones."}],[{"text":"Where the piece sags."}]]},{"t":"p","runs":[{"text":"Watch it once, all the way through, without stopping to fix anything. Take notes on paper. Stopping to fix is the beginning of the hour on shot three."}]},{"t":"h2","id":"cutting-the-length-down","runs":[{"text":"Cutting the length down"}]},{"t":"p","runs":[{"text":"From stringout to rough cut, the operations are almost all subtraction:"}]},{"t":"ul","items":[[{"text":"Trim the head and tail of every clip.","b":true},{"text":" Roughly a quarter-second off the front for the settle, and a second off the back for the drift. This is not a stylistic choice; it is removing the parts where the generation announces itself."}],[{"text":"Cut the shots that were doing nothing.","b":true},{"text":" If removing a shot does not hurt, it was not working."}],[{"text":"Shorten everything.","b":true},{"text":" Almost every rough cut is improved by being ten per cent shorter. Commercial editing runs two to four seconds a shot; your instinct for shot length is probably long."}]]},{"t":"h2","id":"pace-is-a-pattern-not-a-speed","runs":[{"text":"Pace is a pattern, not a speed"}]},{"t":"p","runs":[{"text":"A sequence of shots all the same length reads as mechanical. Vary them: a long establishing shot, three short ones, a longer one to let something land."}]},{"t":"p","runs":[{"text":"The simplest useful pattern is to let the shot after a fast run breathe. Three quick cuts then a held shot gives the audience somewhere to arrive, and it is the difference between energy and noise."}]},{"t":"p","runs":[{"text":"In generative work this pattern has a practical bonus: short shots hide drift, and the one long shot can be the one you generated most carefully."}]},{"t":"h2","id":"keeping-the-decision-record","runs":[{"text":"Keeping the decision record"}]},{"t":"p","runs":[{"text":"At the end of the assembly, your shot list should carry, per shot: the chosen take, the reason, and the intended length. That document is what lets you answer a client’s note about shot nine three weeks later without rewatching everything."}]},{"t":"p","runs":[{"text":"It is also what tells you, honestly, how many generations the project actually took: the number your next quote depends on."}]},{"t":"rule"},{"t":"p","runs":[{"text":"Today: build a stringout of everything you have generated for one project and watch it once, straight through, taking notes on paper without touching the timeline."}]}],"check":{"question":[{"text":"What does a stringout (every selected take, untrimmed, in order) reveal that a partially trimmed rough cut does not?"}],"options":[{"text":[{"text":"The true final duration of the piece, which cannot be estimated once clips have been trimmed to different lengths."}],"correct":false,"tempting":[{"text":"Reverses the relationship: a stringout is always far longer than the finished piece and tells you little about final duration."}]},{"text":[{"text":"Where each generation begins to drift, since the untrimmed tails are still present and can be compared across shots."}],"correct":false,"tempting":[{"text":"A real incidental benefit that belongs to the selects pass, where each take was already watched in full."}]},{"text":[{"text":"How well the shots colour-match, because viewing them at full length under identical conditions makes differences in grade and exposure easier to detect than brief trimmed fragments would."}],"correct":false,"tempting":[{"text":"Colour comparison is genuinely easier on longer clips, and it is a later pass whose decisions do not depend on structure being settled."}]},{"text":[{"text":"Whether the sequence holds together and which shots are missing or unnecessary, before any trimming decisions have foreclosed the structure."}],"correct":true,"tempting":null}],"order":[0,1,2,3],"why":null},"prev":{"slug":"conforming-and-project-setup","title":"Decide the frame rate before you import anything","minutes":9},"next":{"slug":"where-a-cut-hides","title":"Cut on motion, and one frame before the failure","minutes":8}}}