Assembly cut from script or transcript
Build an assembly/radio edit that follows a script or transcript.
20 project-authored proposals · taxonomy v1.1 · 0 new benchmark runs. Verifiers and partial-credit rules below are designs, not executed results.
- Typical applications
- Premiere Pro · DaVinci Resolve · Final Cut Pro
- Class GUI-need label
HY· Hybrid
Task proposals · 20
Expand to inspect the complete original wording and proposed checks. Original proposal text remains English in both interfaces.
R06.04.01.01-T01 Cut a 90-second radio edit from a 25-minute interview transcript
Original instruction
Premiere Pro project fixtures/interview.prproj holds a 25-min interview with its transcript. Build sequence 'Radio_90' (85–95 s) that answers Q1–Q3 from input/brief.md, includes the 5 required quote spans in brief order, and places every cut inside a pause ≥ 80 ms. Use text-based editing or the timeline.
- Platforms
mac· macOSwin· Windows- Applications
- Premiere Pro
- Interfaces
desk· Desktop GUI- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
type· Text & formsclick· Point & clickav· Audio/video over time- Work product
WP05· Time-based media- Conditions
licensed· Licensed pro app- Proposed difficulty
- 4 / 5
Proposed verifier
Parse .prproj (gzip XML) sequence 'Radio_90': duration 85–95 s; ASR on a render aligns the 5 quote spans (word recall ≥ 0.95 each) in order; each cut point lies inside a GT silence ≥ 80 ms (VAD on source); no mid-word cut. Q1–Q3 coverage via GT answer spans.
Proposed partial credit
15% duration; 35% quote spans in order; 30% cuts in silences; 20% Q1–Q3 covered.
R06.04.01.01-T02 Assemble scene 12 from the director's preferred takes
Original instruction
DaVinci Resolve project 'Feature_S12' has 4 takes of each of 9 dialogue lines and input/director_notes.pdf naming the preferred take per line. Build timeline 'S12_Assembly' with one clip per line in script order, using the noted take and trimming each to its line (≤ 6 frames pad).
- Platforms
mac· macOSwin· Windowslin· Linux- Applications
- DaVinci Resolve
- Interfaces
desk· Desktop GUI- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
click· Point & clickdrag· Drag & dropav· Audio/video over time- Work product
WP05· Time-based media- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Export timeline via Resolve API to OTIO: 9 clips in order; each source clip equals the GT take; head/tail padding around the GT line boundaries 0–6 frames; no duplicate lines.
Proposed partial credit
45% correct take per line (5 points each); 25% script order; 20% trims within padding; 10% no extras.
R06.04.01.01-T03 Cut a 30-second spot so every VO sentence has a matching shot
Original instruction
Final Cut Pro library fixtures/Spot.fcpbundle contains a VO track (6 sentences), 40 product shots and an end card. Build project 'Spot_30' exactly 30:00 long: VO as primary, a visually matching product shot under each sentence (e.g. 'waterproof' → splash shot), end card last 2 s.
- Platforms
mac· macOS- Applications
- Final Cut Pro
- Interfaces
desk· Desktop GUI- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
drag· Drag & dropvis· Visual judgmentav· Audio/video over time- Work product
WP05· Time-based media- Conditions
licensed· Licensed pro app- Proposed difficulty
- 4 / 5
Proposed verifier
Export FCPXML: duration exactly 30:00 at 25 fps; for each sentence span the overlapping shot is in the GT acceptable-shot set for that sentence; end card covers 28:00–30:00; VO unaltered (audio hash of render region). Final Cut Pro has no scripting API and any FCPXML written on the VM is detected (file-write log), so the project must be cut in the GUI, not imported.
Proposed partial credit
10% exact duration; 60% sentence-shot matches (10 points each); 15% end card; 15% VO intact.
R06.04.01.01-T04 Build a slate-order stringout of all scene 5 takes
Original instruction
In Premiere Pro, bin 'Day3' has 38 clips; input/slate_log.csv maps clip names to scene/take. Create sequence 'S05_Stringout' containing every scene 5 take in slate order (5A-1, 5A-2, … 5C-4) with 1-second gaps between clips.
- Platforms
mac· macOSwin· Windows- Applications
- Premiere Pro
- Interfaces
desk· Desktop GUI- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
click· Point & clickdrag· Drag & drop- Work product
WP05· Time-based media- Conditions
licensed· Licensed pro app- Proposed difficulty
- 2 / 5
Proposed verifier
Parse .prproj: sequence contains exactly the 14 GT scene-5 clips, ordered per slate sort, full length, gaps 1 s ± 1 frame; no other clips.
Proposed partial credit
40% correct clip set; 35% order; 25% gaps.
R06.04.01.01-T05 Build a documentary assembly from a paper edit
Original instruction
input/paper_edit.docx lists 22 bites as interviewee + source timecode in/out across 6 interviews in Resolve project 'Doc_River'. Build timeline 'Assembly_v1' with those bites in paper-edit order, each trimmed to its timecodes, on V1/A1-A2.
- Platforms
mac· macOSwin· Windowslin· Linux- Applications
- DaVinci Resolve
- Interfaces
desk· Desktop GUI- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
type· Text & formsclick· Point & clickcode· Code & commands- Work product
WP05· Time-based media- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
OTIO export: 22 clips in order, source in/out each within 1 frame of the paper edit; correct source clip per bite; audio linked. Script or GUI routes accepted.
Proposed partial credit
40% correct bites present; 30% order; 30% frame-accurate in/out.
R06.04.01.01-T06 Remove fillers and tighten pauses in a video podcast
Original instruction
Premiere Pro project fixtures/pod_ep14.prproj holds a 12-min single-camera podcast. In sequence 'Ep14_Tight' delete the 47 filler words marked in input/fillers.csv (um, uh, you know) and shorten pauses longer than 1.5 s to 0.5 s, keeping audio and video in sync.
- Platforms
mac· macOSwin· Windows- Applications
- Premiere Pro
- Interfaces
desk· Desktop GUI- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
type· Text & formsclick· Point & clickav· Audio/video over time- Work product
WP05· Time-based media- Conditions
licensed· Licensed pro app- Proposed difficulty
- 3 / 5
Proposed verifier
ASR + forced alignment on render: ≥ 90% of listed fillers absent, no non-filler words removed (word recall vs transcript minus fillers ≥ 0.99); remaining pauses ≤ 0.6 s where GT > 1.5 s; A/V offset 0 on all clips (linked).
Proposed partial credit
40% fillers removed; 20% no content loss; 25% pauses shortened; 15% sync intact.
R06.04.01.01-T07 Assemble a dialogue scene from coverage following script intent
Original instruction
Resolve project 'Short_Diner' has wide, two-shot and two singles for scene 3 plus input/scene3_script.pdf with editor notes: open on the wide, singles for lines 4–9, cut to the reaction single on the line 'You knew?', end on the two-shot. Build 'S03_Assembly' following the script with one clip per line.
- Platforms
mac· macOSwin· Windowslin· Linux- Applications
- DaVinci Resolve
- Interfaces
desk· Desktop GUI- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
drag· Drag & dropvis· Visual judgmentav· Audio/video over time- Work product
WP05· Time-based media- Conditions
- None declared
- Proposed difficulty
- 4 / 5
Proposed verifier
OTIO: first clip from the wide; lines 4–9 from singles with the speaking character in frame (GT); the reaction cut uses the listener single overlapping that line; last clip two-shot; dialogue continuity (ASR word recall ≥ 0.97 vs script, no repeated lines). Resolve scripting (Console and external API) is disabled and any XML/EDL/OTIO/AAF file written on the VM is detected, so the timeline cannot be imported from a script.
Proposed partial credit
15% opening wide; 35% singles on speakers; 30% reaction cut and ending; 20% dialogue continuity.
R06.04.01.01-T08 Use ScriptSync to build an assembly in Avid Media Composer
Original instruction
Avid project 'Pilot_Ep1' has scene 22 takes and input/sc22_script.txt. Import the script, ScriptSync the takes, then build sequence 'Sc22_Assy' using circled takes (per script notes) line by line in script order.
- Platforms
mac· macOSwin· Windows- Applications
- Avid Media Composer
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clickdrag· Drag & droptype· Text & forms- Work product
WP05· Time-based media- Conditions
licensed· Licensed pro apptutorial· Follow docs/tutorial- Proposed difficulty
- 4 / 5
Proposed verifier
Export AAF/ALE + sequence EDL: each line's clip is the circled take; order follows script; script bin shows ScriptSync line linkage (exported script with sync marks). Cuts within 8 frames of GT line boundaries. Media Composer has no scripting access on this seat and any AAF/ALE/EDL written on the VM is detected, so the sequence must be built in the GUI.
Proposed partial credit
25% script imported and synced; 40% circled takes per line; 20% order; 15% cut tightness.
R06.04.01.01-T09 Fix three wrong takes in an AI-generated rough cut
Original instruction
Resolve project 'Promo_IS' has a rough cut generated by IntelliScript from input/promo_script.txt. Director notes say lines 3, 7 and 11 used the wrong take (stumble, boom in frame, wrong emphasis). Replace those three with the best alternative take without shifting other cuts' content.
- Platforms
mac· macOSwin· Windows- Applications
- DaVinci Resolve
- Interfaces
desk· Desktop GUI- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clickdrag· Drag & dropav· Audio/video over timevis· Visual judgment- Work product
WP05· Time-based media- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
OTIO diff vs original: only lines 3, 7, 11 changed; replacements are in GT acceptable-take set (no stumble per ASR, no boom per GT boom mask, emphasis per notes); other clips' source ranges identical; total duration change only from replaced clips.
Proposed partial credit
60% three replacements correct (20 points each); 30% other cuts unchanged; 10% no gaps introduced.
R06.04.01.01-T10 Assemble a Spanish-language interview to an English brief
Original instruction
Premiere Pro project fixtures/ES_interview.prproj has a 40-min Spanish interview and its Spanish transcript. input/brief_en.md asks for a 2-minute (115–125 s) piece covering three topics (housing, transit, schools) in that order, at least 30 s per topic.
- Platforms
mac· macOSwin· Windows- Applications
- Premiere Pro
- Interfaces
desk· Desktop GUI- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
type· Text & formsclick· Point & clickav· Audio/video over time- Work product
WP05· Time-based media- Conditions
licensed· Licensed pro app- Proposed difficulty
- 4 / 5
Proposed verifier
Render + Spanish ASR; GT topic span labels on source. Duration 115–125 s; ≥ 30 s from each topic; topic order housing→transit→schools; cuts at pauses ≥ 80 ms; no off-topic segment longer than 3 s.
Proposed partial credit
15% duration; 40% topic coverage; 20% order; 25% clean cuts and no off-topic.
R06.04.01.01-T11 Build a 60-second customer testimonial from three interviews
Original instruction
Final Cut Pro library fixtures/Testimonials.fcpbundle has interviews with speakers A, B, C and transcripts. Build project 'Testimonial_60' (58–62 s) in speaker order A, B, C, A using bites marked as favorites, with no cut inside a word.
- Platforms
mac· macOS- Applications
- Final Cut Pro
- Interfaces
desk· Desktop GUI- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clickdrag· Drag & dropav· Audio/video over time- Work product
WP05· Time-based media- Conditions
licensed· Licensed pro app- Proposed difficulty
- 3 / 5
Proposed verifier
FCPXML: duration 58–62 s; speaker sequence of clips A,B,C,A (merging consecutive same-speaker clips); all source ranges inside favorite ranges; cut points within GT inter-word gaps.
Proposed partial credit
20% duration; 35% speaker order; 25% favorites only; 20% clean cut points.
R06.04.01.01-T12 Build a news package A-roll from a script with SOTs
Original instruction
input/package_script.txt alternates VO paragraphs and SOT bites (with in-words/out-words). In Premiere Pro, lay the recorded VO from bin 'VO' on A1 and cut each SOT (picture + sound) from bin 'Interviews' at its in/out words, leaving VO sections' picture as black slugs for b-roll.
- Platforms
mac· macOSwin· Windows- Applications
- Premiere Pro
- Interfaces
desk· Desktop GUI- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
type· Text & formsdrag· Drag & dropav· Audio/video over time- Work product
WP05· Time-based media- Conditions
licensed· Licensed pro app- Proposed difficulty
- 3 / 5
Proposed verifier
.prproj: VO segments in script order with no overlap; each SOT's source range starts/ends at the GT in/out words (±6 frames); V1 empty or black slug under VO; total = sum of parts with 0–10 frame gaps.
Proposed partial credit
30% VO order; 40% SOT in/out accuracy; 15% slugs; 15% no gaps/overlaps.
R06.04.01.01-T13 Clean a lecture recording using transcript notes
Original instruction
Resolve project 'Lecture_07' has a 75-min single-take lecture. input/edit_notes.md flags 12 off-topic segments to remove and 3 false starts where only the final attempt stays. Produce timeline 'Lecture_07_Clean' and keep chapter markers at the 6 topic starts listed.
- Platforms
mac· macOSwin· Windowslin· Linux- Applications
- DaVinci Resolve
- Interfaces
desk· Desktop GUI- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
type· Text & formsclick· Point & clickav· Audio/video over time- Work product
WP05· Time-based media- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
OTIO: removed ranges match GT off-topic spans (IoU ≥ 0.9 total); each false start leaves only the last attempt; 6 markers within 2 s of the topic starts in the new timeline; no other content removed (> 99% of kept GT duration present).
Proposed partial credit
40% off-topic removal; 20% false starts; 25% chapter markers; 15% nothing else lost.
R06.04.01.01-T14 Organize transcript topics with keywords and assemble in topic order
Original instruction
In Final Cut Pro, event 'Founders' has 5 interviews with transcripts. Keyword ranges for topics 'Origin', 'Struggle', 'Breakthrough' using input/topic_spans.csv as guide, then build project 'Founders_Story' with one clip per keyword range ordered Origin → Struggle → Breakthrough.
- Platforms
mac· macOS- Applications
- Final Cut Pro
- Interfaces
desk· Desktop GUI- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
click· Point & clickdrag· Drag & droptype· Text & forms- Work product
WP05· Time-based media- Conditions
licensed· Licensed pro app- Proposed difficulty
- 3 / 5
Proposed verifier
FCPXML: keyword ranges exist on clips with names and ranges within 1 s of CSV spans; project contains those ranges grouped by topic order; no unrelated clips.
Proposed partial credit
40% keyword ranges; 40% project order and coverage; 20% no unrelated clips.
R06.04.01.01-T15 Assemble a reality-show story from logger notes and footage
Original instruction
Premiere Pro project fixtures/Reality_E3.prproj holds 2 h of multicam-free footage (12 clips) and input/story_beats.md (8 beats) with logger notes. Build a 6-minute (±15 s) assembly 'E3_ActB' that tells the 8 beats in order, each beat sourced from the clip moments the notes describe, using only on-camera dialogue.
- Platforms
mac· macOSwin· Windows- Applications
- Premiere Pro
- Interfaces
desk· Desktop GUI- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
av· Audio/video over timedrag· Drag & droptype· Text & formsvis· Visual judgment- Work product
WP05· Time-based media- Conditions
licensed· Licensed pro applong· Long-horizon- Proposed difficulty
- 5 / 5
Proposed verifier
.prproj: duration 5:45–6:15; for each beat ≥ 1 clip overlapping its GT source window (logged moments); beat order preserved; ASR shows no orphan half-sentences at cuts (cut points in pauses). VLM not used.
Proposed partial credit
10% duration; 50% beats covered (6.25 points each); 20% beat order; 20% clean cuts.
R06.04.01.01-T16 Control: generate the T01 radio edit as OTIO from the transcript
Original instruction
Control pair of T01 (CLI half). From input/interview_transcript.json (word timings) and input/brief.md, write a script that outputs out/radio_90.otio (85–95 s) containing the 5 required quote spans in order with cuts inside pauses ≥ 80 ms; render it with ffmpeg to out/radio_90.wav.
- Platforms
lin· Linuxmac· macOSwin· Windows- Applications
- Python · OpenTimelineIO · ffmpeg
- Interfaces
term· Terminal- GUI need · task setting
CL· CLI-sufficient- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
code· Code & commands- Work product
WP05· Time-based media- Conditions
- None declared
- Proposed difficulty
- 2 / 5
Proposed verifier
Same metrics as T01 on the OTIO and WAV. Paired with T01 to compare GUI text-based editing against a pure transcript-to-timeline script.
Proposed partial credit
15% duration; 35% quote spans; 30% cuts in pauses; 20% Q1–Q3 coverage.
R06.04.01.01-T17 Assemble a short interview in Kdenlive from a transcript
Original instruction
Kdenlive project fixtures/maker.kdenlive has a 15-min interview and input/transcript.srt. Build a 3-minute (175–185 s) cut on the main timeline containing the 4 answers highlighted in input/selects.md in order, removing the interviewer's questions.
- Platforms
lin· Linuxmac· macOSwin· Windows- Applications
- Kdenlive
- Interfaces
desk· Desktop GUI- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
drag· Drag & dropclick· Point & clickav· Audio/video over time- Work product
WP05· Time-based media- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Parse .kdenlive (MLT XML): duration 175–185 s; each highlighted answer's span present (coverage ≥ 95%) in order; interviewer speech (GT diarization) ≤ 1 s total; cuts in pauses.
Proposed partial credit
15% duration; 45% answers covered; 15% order; 25% questions removed and clean cuts.
R06.04.01.01-T18 Build a selects reel of every mention of one keyword
Original instruction
In Premiere Pro project fixtures/climate_series.prproj (8 interviews with transcripts), search transcripts for every utterance containing 'drought' and build sequence 'Drought_Selects' with each full sentence plus 12-frame handles, in interview order.
- Platforms
mac· macOSwin· Windows- Applications
- Premiere Pro
- Interfaces
desk· Desktop GUI- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
type· Text & formsclick· Point & click- Work product
WP05· Time-based media- Conditions
licensed· Licensed pro app- Proposed difficulty
- 2 / 5
Proposed verifier
.prproj: clips cover all 17 GT sentences (recall ≥ 0.95) and no unrelated sentences (precision ≥ 0.9); each clip extends 12 ± 2 frames beyond sentence bounds; interview order preserved.
Proposed partial credit
50% recall; 20% precision; 20% handles; 10% order.
R06.04.01.01-T19 Assemble a feature scene from a lined script and circled takes
Original instruction
Resolve project 'Feature_Sc41' has 26 takes across 5 setups and input/lined_script.pdf (coverage lines, circled takes, continuity notes: cup in right hand from line 6). Build 'Sc41_Assy' covering every line once with circled takes, honoring the continuity note and placing the wild line WT-3 over the wide.
- Platforms
mac· macOSwin· Windowslin· Linux- Applications
- DaVinci Resolve
- Interfaces
desk· Desktop GUI- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
drag· Drag & dropvis· Visual judgmentav· Audio/video over timetype· Text & forms- Work product
WP05· Time-based media- Conditions
long· Long-horizon- Proposed difficulty
- 5 / 5
Proposed verifier
OTIO: every script line covered exactly once (ASR alignment); each clip is a circled take for a setup covering that line per lined script; from line 6 onward no clip with cup in left hand (GT per-take flags); WT-3 audio placed under the wide clip. Resolve scripting (Console and external API) is disabled and any XML/EDL/OTIO/AAF file written on the VM is detected, so the timeline cannot be imported from a script.
Proposed partial credit
35% full line coverage; 30% circled takes/coverage; 20% continuity respected; 15% wild line placement.
R06.04.01.01-T20 Conform an assembly to a revised blue-page script
Original instruction
Premiere Pro sequence 'Ep5_Assy' was cut to the white script. input/script_blue.pdf revises scenes 14–16: 4 lines deleted, 2 lines changed (pickups in bin 'Pickups'), scene 15 moved after 16. Update the sequence to match the blue pages, keeping untouched sections frame-identical.
- Platforms
mac· macOSwin· Windows- Applications
- Premiere Pro
- Interfaces
desk· Desktop GUI- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
drag· Drag & droptype· Text & formsav· Audio/video over time- Work product
WP05· Time-based media- Conditions
licensed· Licensed pro applong· Long-horizon- Proposed difficulty
- 5 / 5
Proposed verifier
.prproj diff: deleted lines absent (ASR); changed lines use pickup clips; scene order 14,16,15; all clips outside affected lines keep identical source ranges and relative order; no gaps.
Proposed partial credit
25% deletions; 25% pickups placed; 25% scene reorder; 25% untouched sections identical.
Mapped research references
Existing benchmark-to-class links from the research catalog. They are not record-level annotations, executions or product endorsements.