Short-form & template editing
Edit with templates, themes and trailers (iMovie, CapCut) and cut vertical short videos with captions and effects.
20 project-authored proposals · taxonomy v1.1 · 0 new benchmark runs. Verifiers and partial-credit rules below are designs, not executed results.
- Typical applications
- CapCut · Jianying · Premiere Rush
- Class GUI-need label
G3· GUI-only
Task proposals · 20
Expand to inspect the complete original wording and proposed checks. Original proposal text remains English in both interfaces.
R06.04.01.04-T01 Reframe a landscape clip into a 30-second vertical short with captions
Original instruction
In CapCut desktop, import fixtures/cooking_16x9.mp4 (2 min). Make a 9:16 1080×1920 short of 28–32 s from the segments listed in input/selects.txt, keep the chef's hands inside the frame (keyframe the crop), add auto captions corrected to input/transcript.txt, and export out/cooking_short.mp4.
- Platforms
mac· macOSwin· Windows- Applications
- CapCut
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
drag· Drag & dropgeom· Precise geometrytype· Text & formsvis· Visual judgment- Work product
WP05· Time-based media- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Export 1080×1920; duration 28–32 s; content matches selects (frame fingerprint); GT hands bbox inside frame on ≥ 95% frames; OCR captions vs transcript WER ≤ 5%. Export must carry CapCut encoder metadata and a matching draft in the app's draft folder (ffmpeg-only renders fail). ffmpeg is absent from the VM and draft files written by other processes are detected.
Proposed partial credit
15% format/duration; 20% correct segments; 35% subject kept in frame; 30% captions accurate.
R06.04.01.04-T02 Fill a CapCut mobile template with six clips in order
Original instruction
On an Android phone, in CapCut open template 'Weekend Recap 6' (saved in the test account's favorites), replace its 6 placeholders with fixtures/recap/clip1–clip6.mp4 in numeric order, adjust each clip's visible segment to the moment listed in input/moments.csv, and export 1080p to the gallery as recap.mp4.
- Platforms
and· Android- Applications
- CapCut
- Interfaces
mob· Mobile touch- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G3· GUI-only- Interaction primitives
touch· Touch gesturesdrag· Drag & dropvis· Visual judgment- Work product
WP05· Time-based media- Conditions
- None declared
- Proposed difficulty
- 2 / 5
Proposed verifier
Exported file in DCIM: 6 template slots detected by template timing; each slot's content matches the right clip (fingerprint) and its segment contains the listed moment; 1080p output. adb shell access is disabled and the phone has no ffmpeg, so the video can only be assembled in the CapCut app.
Proposed partial credit
40% clips in correct order; 40% segments include moments; 20% export spec.
R06.04.01.04-T03 Correct Chinese auto-captions and style them in Jianying
Original instruction
In Jianying (剪映) desktop, import fixtures/vlog_zh.mp4 (3 min). Generate auto captions, correct them to input/transcript_zh.txt, style: white text, black stroke, 46 px, bottom-centered with baseline at 85% frame height, max 16 characters per line. Export out/vlog_zh_sub.mp4.
- Platforms
mac· macOSwin· Windows- Applications
- Jianying
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- 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
OCR on sampled frames vs transcript CER ≤ 2%; caption timing overlaps GT speech segments ≥ 90%; style: text color white with dark stroke, glyph height in range, baseline within 2% of 85% height; no line > 16 chars. Jianying export metadata required. ffmpeg is absent from the VM and draft_content.json writes by other processes (e.g. pyJianYingDraft) are detected; the export must carry Jianying encoder metadata.
Proposed partial credit
45% caption text accuracy; 20% timing; 25% style/position; 10% line length.
R06.04.01.04-T04 Make a beat-synced photo slideshow with CapCut's beat markers
Original instruction
In CapCut desktop, create a 1080×1920 slideshow from the 12 images in fixtures/trip_photos/ over fixtures/song.mp3 (first 24 s). Use the beat detection markers so each photo change lands on a beat (two beats per photo), add a slow zoom on each photo, export out/trip_slideshow.mp4.
- Platforms
mac· macOSwin· Windows- Applications
- CapCut
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
drag· Drag & dropav· Audio/video over timeclick· Point & click- Work product
WP05· Time-based media- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Scene-change detection on export: 12 photo segments in folder order; each change within ±1 frame of GT beats at a 2-beat spacing; per-segment scale increase 3–15% (zoom present); audio is the song's first 24 s. ffmpeg is absent from the VM and draft files written by other processes are detected, so the export must come from CapCut (encoder metadata plus matching app draft).
Proposed partial credit
20% 12 photos in order; 45% changes on beats; 20% zoom on each; 15% audio/format.
R06.04.01.04-T05 Key a green-screen presenter over a background in CapCut
Original instruction
In CapCut desktop, place fixtures/presenter_green.mp4 over fixtures/office_bg.jpg in a 1080×1920 project, remove the green with Chroma key (adjust intensity and shadow), position the presenter bottom-left occupying 60% of height, and export out/presenter.mp4.
- Platforms
mac· macOSwin· Windows- Applications
- CapCut
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
click· Point & clickgeom· Precise geometryvis· Visual judgment- Work product
WP05· Time-based media- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Per sampled frame: alpha-derived presenter mask vs GT matte IoU ≥ 0.95; green spill on edge band (G − max(R,B) mean) ≤ 0.03; presenter bbox height 58–62% of frame, anchored bottom-left within 40 px; background visible elsewhere. ffmpeg is absent from the VM and draft files written by other processes are detected, so the export must come from CapCut (encoder metadata plus matching app draft).
Proposed partial credit
45% key quality; 20% spill; 25% placement; 10% export spec.
R06.04.01.04-T06 Build a 15-second product ad respecting TikTok safe zones
Original instruction
In CapCut web, using the brand kit in input/brand_kit.pdf (2 hex colors, logo PNG, font), make a 15 s 9:16 ad from fixtures/product_clips/: hook text in first 2 s, 3 feature captions, price card, logo end card for the last 2 s. No text may enter the bottom 20% or right 15% (platform UI). Add music ducked under the VO. Export out/ad15.mp4.
- Platforms
web· Any OS (web)- Applications
- CapCut
- Interfaces
brow· Browser- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
drag· Drag & droptype· Text & formsgeom· Precise geometryvis· Visual judgment- Work product
WP05· Time-based media- Conditions
long· Long-horizon- Proposed difficulty
- 5 / 5
Proposed verifier
Duration 15.0 s; OCR finds hook (0–2 s), 3 features, price; all text bboxes outside unsafe zones; text colors within ΔE2000 ≤ 3 of brand hex; logo match in last 2 s; music −10 dB or more under VO segments (LUFS short-term). The test CapCut web account has no API token and ffmpeg is absent from the VM, so the export must be a download from the browser editor session.
Proposed partial credit
35% structure/timing and brand colors/font; 25% safe zones respected; 20% end card logo; 20% audio ducking.
R06.04.01.04-T07 Add keyframed punch-in zooms and word-timed text pops in Jianying
Original instruction
Jianying desktop project fixtures/talk.draft (talking head). Add keyframed 120% punch-in zooms at the 3 emphasis moments in input/emphasis.csv (ramp over 6 frames, hold, return), and pop-in text for the 3 keywords timed to when each word is spoken (±3 frames). Export out/talk_punch.mp4.
- Platforms
mac· macOSwin· Windows- Applications
- Jianying
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G3· GUI-only- Interaction primitives
geom· Precise geometrydrag· Drag & dropav· Audio/video over time- Work product
WP05· Time-based media- Conditions
- None declared
- Proposed difficulty
- 4 / 5
Proposed verifier
Face-bbox scale from export: 3 zoom events at listed times, peak 115–125%, ramp 4–8 frames; outside events scale 100% ± 2%; OCR keyword appearance within ±3 frames of GT word onset. ffmpeg is absent from the VM and draft_content.json writes by other processes (e.g. pyJianYingDraft) are detected; the export must carry Jianying encoder metadata.
Proposed partial credit
45% three zooms correct; 15% ramp lengths; 40% keyword pops timed.
R06.04.01.04-T08 Tighten a talking-head short and overlay B-roll at keywords
Original instruction
In CapCut desktop, import fixtures/coach_raw.mp4 (90 s). Remove silences longer than 0.4 s (jump cuts allowed), then overlay the 4 B-roll clips from fixtures/broll/ as picture overlays when the matching keywords in input/keywords.csv are spoken, each 2 s. Export out/coach_tight.mp4.
- Platforms
mac· macOSwin· Windows- Applications
- CapCut
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
drag· Drag & dropav· Audio/video over timeclick· Point & click- Work product
WP05· Time-based media- Conditions
- None declared
- Proposed difficulty
- 4 / 5
Proposed verifier
VAD on export: no pause > 0.45 s; ASR word recall vs source transcript ≥ 0.99; each B-roll visible (fingerprint) starting within ±6 frames of its keyword onset, duration 2 ± 0.1 s, full-frame overlay. ffmpeg is absent from the VM and draft files written by other processes are detected, so the export must come from CapCut (encoder metadata plus matching app draft).
Proposed partial credit
35% silences removed; 15% no words lost; 40% B-roll timing (10 points each); 10% export.
R06.04.01.04-T09 Edit a music-timed reel in Instagram's Edits app
Original instruction
On an iPhone, in the Edits app, build a 20 s reel from the 5 clips in the album 'Hike' using the audio fixtures/hike_track.m4a, cutting on the 5 beat markers listed in input/beats.txt, and add the text 'Summit day' from 1 s to 4 s. Save to Photos as hike_reel.mp4 (do not post).
- Platforms
ios· iOS/iPadOS- Applications
- Instagram Edits
- Interfaces
mob· Mobile touch- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
touch· Touch gesturesdrag· Drag & dropav· Audio/video over time- Work product
WP05· Time-based media- Conditions
risk· Irreversible/risky- Proposed difficulty
- 3 / 5
Proposed verifier
Video in Photos: 20 s; 5 segments in album order; cuts within ±2 frames of listed beats; OCR 'Summit day' visible 1–4 s only; no post made (test account activity feed empty). The iPhone offers no shell or ffmpeg and the test account has no Graph API token, so the reel can only be edited in the Edits app.
Proposed partial credit
20% length/clips; 40% cuts on beats; 25% text timing; 15% saved not posted.
R06.04.01.04-T10 Assemble a quick title-plus-clips video in Premiere Rush on Android
Original instruction
On an Android tablet, in Premiere Rush, create a project from fixtures/market/ clips 1–4 in order, trim each to 4 s, add the title 'Night Market' over the first clip, apply the 'Warm' color preset to all clips, and export 1080p to Downloads/market.mp4.
- Platforms
and· Android- Applications
- Premiere Rush
- Interfaces
mob· Mobile touch- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G3· GUI-only- Interaction primitives
touch· Touch gesturestype· Text & forms- Work product
WP05· Time-based media- Conditions
- None declared
- Proposed difficulty
- 2 / 5
Proposed verifier
Export: 16 s ± 2 frames; 4 segments in order (fingerprint); OCR title in first clip; mean b* of each clip higher than source by ≥ 3 (warm preset); 1080p. adb shell access is disabled and the tablet has no ffmpeg, so the video can only be assembled in Premiere Rush.
Proposed partial credit
35% clips trimmed and ordered; 25% title; 25% color preset; 15% export spec.
R06.04.01.04-T11 Speed-ramp a skate trick into slow motion at its peak
Original instruction
In CapCut desktop apply a custom speed curve to fixtures/kickflip.mp4 (240 fps source): 2× speed before takeoff, 0.3× around the peak (GT frames 1820–1940), 2× after landing, with smooth transitions; final clip 6–8 s. Export out/kickflip_ramp.mp4 at 1080×1920 with the skater centered.
- Platforms
mac· macOSwin· Windows- Applications
- CapCut
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
geom· Precise geometrydrag· Drag & dropvis· Visual judgment- Work product
WP05· Time-based media- Conditions
- None declared
- Proposed difficulty
- 4 / 5
Proposed verifier
Frame-to-source mapping via fingerprints: speed ≈ 2× (±0.2) outside, 0.3× (±0.05) in peak window; transitions without speed jumps > 0.5× per frame; duration 6–8 s; skater GT bbox center within the middle 40% width on ≥ 90% frames. ffmpeg is absent from the VM and draft files written by other processes are detected, so the export must come from CapCut (encoder metadata plus matching app draft).
Proposed partial credit
40% peak slow-mo; 20% fast sections; 15% smooth curve; 25% framing and duration.
R06.04.01.04-T12 Publish a reusable five-slot template in CapCut
Original instruction
In CapCut desktop, turn project fixtures/promo_base into a template with 5 replaceable media slots (the 5 product shots) and 2 editable text fields (product name, price), locking music and transitions. Save it to the test account as 'Promo5' (private, do not publish publicly) and export a preview using input/sample_media/.
- Platforms
mac· macOSwin· Windows- Applications
- CapCut
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G3· GUI-only- Interaction primitives
click· Point & clicktype· Text & formsdrag· Drag & drop- Work product
WP08· App/system state- Conditions
risk· Irreversible/risky- Proposed difficulty
- 3 / 5
Proposed verifier
Test account template list via the app: 'Promo5' private with 5 media slots and 2 text fields; preview export shows sample media in the 5 slots (fingerprint) and sample text (OCR); music identical to base; visibility not public. CapCut issues no API token for the test account and draft files written by other processes are detected, so the template must be built in the app.
Proposed partial credit
30% 5 media slots; 20% 2 text fields; 20% locked music/transitions; 30% private save and preview correct.
R06.04.01.04-T13 Produce Spanish and Japanese caption versions with identical timing
Original instruction
CapCut desktop project fixtures/tips_en (45 s, English captions). Duplicate it into 'tips_es' and 'tips_ja', replace caption text with input/captions_es.srt and captions_ja.srt content while keeping each caption's timing, adjust font for Japanese glyphs, and export all three to out/.
- Platforms
mac· macOSwin· Windows- Applications
- CapCut
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G3· GUI-only- Interaction primitives
type· Text & formsclick· Point & click- Work product
WP05· Time-based media- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Three exports; OCR per version vs its SRT CER ≤ 3%; caption on/off times identical across versions (±1 frame); Japanese glyphs render (no tofu boxes) and stay inside the safe area. ffmpeg is absent from the VM and draft files written by other processes are detected, so the export must come from CapCut (encoder metadata plus matching app draft).
Proposed partial credit
25% English export unchanged; 30% Spanish correct; 30% Japanese correct; 15% timing identical.
R06.04.01.04-T14 Track a price label to a moving product in Jianying
Original instruction
In Jianying desktop, attach the text sticker '¥199' to the sneaker in fixtures/shoe_spin.mp4 using motion tracking for its full 5 s, so the label stays above the shoe as it moves. Export out/shoe_label.mp4.
- Platforms
mac· macOSwin· Windows- Applications
- Jianying
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
geom· Precise geometryclick· Point & clickvis· Visual judgment- Work product
WP05· Time-based media- Conditions
- None declared
- Proposed difficulty
- 4 / 5
Proposed verifier
Per frame: OCR '¥199' present; label center within 25 px horizontally of GT shoe bbox center and 40–120 px above its top edge on ≥ 95% frames; label never overlaps the shoe bbox. ffmpeg is absent from the VM and draft_content.json writes by other processes (e.g. pyJianYingDraft) are detected; the export must carry Jianying encoder metadata.
Proposed partial credit
20% label present; 55% tracking accuracy; 25% no overlap.
R06.04.01.04-T15 Match eight vlog clips to a reference look with filters and adjustments
Original instruction
In CapCut desktop, project fixtures/beach_vlog has 8 clips with mixed white balance. Using filters and the Adjust panel per clip, match each clip to input/reference_frame.png look: neutral gray patch, skin tone and sky hue. Export out/beach_matched.mp4.
- Platforms
mac· macOSwin· Windows- Applications
- CapCut
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
click· Point & clickvis· Visual judgment- Work product
WP05· Time-based media- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Per clip mid-frame: GT neutral patch |a*|,|b*| ≤ 3; skin mask mean ΔE2000 to reference ≤ 5; sky hue within 8° of reference; clip-to-clip skin ΔE2000 ≤ 4. ffmpeg is absent from the VM and draft files written by other processes are detected, so the export must come from CapCut (encoder metadata plus matching app draft).
Proposed partial credit
35% neutrals; 30% skin match; 20% sky hue; 15% consistency across clips.
R06.04.01.04-T16 Lay out a reaction video with picture-in-picture and ducking
Original instruction
In CapCut desktop, build a 1920×1080 reaction video: fixtures/trailer.mp4 full-frame, fixtures/reaction_cam.mp4 as picture-in-picture top-right at 30% width with 24 px margin and rounded corners, synced so the reaction clap aligns with the trailer's explosion (frame given), and the trailer audio ducked −12 dB whenever the reactor speaks.
- Platforms
mac· macOSwin· Windows- Applications
- CapCut
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G3· GUI-only- Interaction primitives
geom· Precise geometryav· Audio/video over timedrag· Drag & drop- Work product
WP05· Time-based media- Conditions
- None declared
- Proposed difficulty
- 4 / 5
Proposed verifier
PIP bbox width 576 ± 6 px, right/top margin 24 ± 3 px; corner radius > 0 (alpha check); sync: clap frame within ±1 of explosion; trailer audio level −12 ± 2 dB during GT reactor speech vs elsewhere. ffmpeg is absent from the VM and draft files written by other processes are detected, so the export must come from CapCut (encoder metadata plus matching app draft).
Proposed partial credit
30% PIP geometry; 10% rounded corners; 30% sync; 30% ducking.
R06.04.01.04-T17 Make a 50/50 vertical split-screen duet
Original instruction
In CapCut on iPhone, create a 1080×1920 split screen: fixtures/teacher.mp4 on top half and fixtures/student.mp4 on bottom half, each scaled to fill its half without black bars, audio of both kept, and start aligned on the shared count-in clap. Save to Photos as duet.mp4.
- Platforms
ios· iOS/iPadOS- Applications
- CapCut
- Interfaces
mob· Mobile touch- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G3· GUI-only- Interaction primitives
touch· Touch gesturesgeom· Precise geometryav· Audio/video over time- Work product
WP05· Time-based media- Conditions
- None declared
- Proposed difficulty
- 2 / 5
Proposed verifier
Frame analysis: top half shows teacher, bottom half student (fingerprint), no black bars (> 2% black rows); clap transients aligned within 1 frame; both audio sources present (cross-correlation). The iPhone offers no shell or ffmpeg, so the split screen can only be built in the CapCut app.
Proposed partial credit
40% split layout filled; 35% sync on clap; 25% both audio tracks.
R06.04.01.04-T18 Produce a 60-second explainer short from a script
Original instruction
In CapCut desktop, produce a 60 s (±1 s) 9:16 explainer from input/script.md (8 scenes) using fixtures/vo.wav, stock clips in fixtures/stock/, and the brand font. Each scene's visual must match its line (e.g. 'solar panels' → panel shot), captions burned, transitions only between scenes, music at −20 LUFS under VO, CTA end card.
- Platforms
mac· macOSwin· Windows- Applications
- CapCut
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
drag· Drag & droptype· Text & formsav· Audio/video over timevis· Visual judgment- Work product
WP05· Time-based media- Conditions
long· Long-horizon- Proposed difficulty
- 5 / 5
Proposed verifier
Duration 59–61 s; per scene, shot under each VO line is in GT acceptable set; caption OCR WER ≤ 5%; transitions only at scene boundaries; music −20 ± 2 LUFS during VO gaps and ducked under VO; end card OCR CTA in last 3 s. ffmpeg is absent from the VM and draft files written by other processes are detected, so the export must come from CapCut (encoder metadata plus matching app draft).
Proposed partial credit
20% duration and transitions; 35% scene visuals; 20% captions; 25% audio levels and CTA end card.
R06.04.01.04-T19 Export five hook variants of the same ad for A/B testing
Original instruction
CapCut desktop project fixtures/ad_body (12 s body). Create 5 versions, each with a different 3 s hook from input/hooks.csv (clip + on-screen text) prepended, body identical, and export out/ad_hook1.mp4 … ad_hook5.mp4 at 1080×1920, 30 fps.
- Platforms
mac· macOSwin· Windows- Applications
- CapCut
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G3· GUI-only- Interaction primitives
drag· Drag & droptype· Text & formsclick· Point & click- Work product
WP05· Time-based media- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
5 files, each 15 s; first 3 s contain the right hook clip (fingerprint) and hook text (OCR); remaining 12 s frame-identical across files (perceptual hash); 30 fps, 1080×1920. ffmpeg is absent from the VM and draft files written by other processes are detected, so the export must come from CapCut (encoder metadata plus matching app draft).
Proposed partial credit
50% five hooks correct (10 points each); 30% body identical; 20% export spec and naming.
R06.04.01.04-T20 Produce a Douyin product video with timed price stickers and loudness
Original instruction
In Jianying desktop build a 30 s 9:16 1080p60 e-commerce video from fixtures/kettle_clips/: burned-in Chinese subtitles from input/vo_zh.srt, price sticker '限时 ¥129' appearing exactly when the VO says the price, 3 feature callouts with arrows pointing at the feature, BGM at −14 LUFS integrated overall. Export out/kettle_douyin.mp4.
- Platforms
mac· macOSwin· Windows- Applications
- Jianying
- Interfaces
desk· Desktop GUI- GUI need · task setting
G3· GUI-only- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
drag· Drag & droptype· Text & formsgeom· Precise geometryav· Audio/video over time- Work product
WP05· Time-based media- Conditions
long· Long-horizon- Proposed difficulty
- 5 / 5
Proposed verifier
1080×1920 at 60 fps, 30 s; subtitle OCR CER ≤ 3%; sticker onset within ±3 frames of GT price word; 3 arrows' tips within 40 px of GT feature points; integrated loudness −14 ± 1 LUFS, true peak ≤ −1 dBTP. ffmpeg is absent from the VM and draft_content.json writes by other processes (e.g. pyJianYingDraft) are detected; the export must carry Jianying encoder metadata.
Proposed partial credit
35% format and subtitles; 20% price sticker timing; 25% callout arrows; 20% loudness.
Mapped research references
Existing benchmark-to-class links from the research catalog. They are not record-level annotations, executions or product endorsements.