Editing practice / 30 fps / native tools
Editing with
DaVinci Resolve
From a visible editing mechanic to a named timeline, an explicit frame budget and a construction plan. Source evidence and implementation choices stay separate.
Choose by editorial intent
What does this beat need?
A hard cut, a push and a masked scroll solve different problems. Use one only when it advances the recorded argument.
| Intent | Recipe | Chosen duration | Page |
|---|---|---|---|
| Introduce | Title hold → hard cutR01_title_cut | 150 f / 5 s | Edit |
| Structure | Chapter reset → return to contextR02_chapter_reset | 180 f / 6 s | Edit |
| Focus | Context → detail by hard cutR03_context_detail | 120 f / 4 s | Edit |
| Focus | Push, settle, then actR04_camera_push | 120 f / 4 s | Edit |
| Show cause | Target → dwell → inputR05_pointer_action | 180 f / 6 s | Edit |
| Show cause | Keep a selected scope through a reframeR06_selection_scope | 150 f / 5 s | Fusion |
| Navigate | Scroll content behind a fixed viewportR07_masked_scroll | 150 f / 5 s | Fusion |
| Prove | Reveal a container, then its contentsR08_result_reveal | 150 f / 5 s | Fusion |
| Prove | Protect a complete-result reading holdR09_reading_hold | 120 f / 4 s | Edit |
| Connect | Carry identity across a context cutR10_object_continuity | 150 f / 5 s | Edit |
| Label | Keep an overlay independent of the shotR11_anchored_overlay | 150 f / 5 s | Edit |
| End | Finish proof, then release attentionR12_quiet_ending | 120 f / 4 s | Edit |
Evidence
The film is the observation.
Each recipe links the official upload, publication date, inspected interval, source-frame cadence and annotated boundaries. Short silent source excerpts remain separate from original assets.
Observed pixels and PTS are measurements. Editorial purpose is interpretation. Geometry, linear interpolation, colors and reading durations below are chosen defaults—not recovered OpenAI settings.
Implementation
The timeline is still to be tested.
Instructions are checked against the official Resolve 21 manual. No available Resolve application means no verified node behavior, project interchange, output codec or rendered fidelity. The guide makes those gaps visible rather than packaging a fake project.
Sequence structure and pacing
One original, silent 52-second study.
This is a 1560-frame assembly specification, not a rendered example. Copy native recipe layers into S01_Original_Study_52s; never substitute flattened videos. In particular, result arrival continues into 6.5 seconds of complete-result reading time.
| Start frame | Duration | Beat | Construction note |
|---|---|---|---|
| 0 | 90 f | PREMISE | Title portion only, recipe frames 0–89; cut is the illustrative content cue, not an actual OBS offset. |
| 90 | 60 f | CUE_CONTEXT | Context portion only, recipe frames 90–149; cue at study frame 90 = 3 seconds. No talking-head recording is paired. |
| 150 | 120 f | FOCUS | Full recipe. Ends on enlarged context. |
| 270 | 150 f | ACTION | Use frames 30–179: cursor travels immediately, then dwells, then input builds; last 60 frames remain intact. |
| 420 | 180 f | CHAPTER | Full recipe. Direct context change is deliberate, not object continuity. |
| 600 | 150 f | SCOPE | Full recipe; selection stays with document during the outer reframe. |
| 750 | 150 f | NAVIGATE | Full recipe. Cut to unhighlighted document is a new navigation beat. |
| 900 | 120 f | RESULT | Use frames 0–119; complete result appears at local 75. Continue identical final state in the next beat. |
| 1020 | 150 f | READ_START | Extend recipe to 150 frames. Combined with preceding complete-result tail, total result reading hold is 195 frames = 6.5 seconds. |
| 1170 | 150 f | IDENTITY | Full recipe. Returning to the original request is an explicit recap in another context. |
| 1320 | 120 f | LABEL | Use frames 15–134; overlay lasts for this whole excerpt, cut occurs at assembly-local frame 45. |
| 1440 | 120 f | ENDPOINT | Full recipe. No source logo, no fade to black, no recording buffer assumed. |
Do not transplant these durations onto a speech recording. Derive a timed transcript from the actual footage, schedule proof against speech anchors, then protect action and reading intervals.
Selection and limits
Representative coverage, not a census.
Publication window: September 5, 2025–September 5, 2026, inclusive. The inherited watch-page metadata inventory contains 421 in-period records; this chapter uses nine films and twelve bounded windows. Collection cutoff: 2026-09-05T01:21:33Z. Later uploads on the final day are unknown.
Dates come from channel-checked watch-page publication metadata, not download dates. The earlier 100-film collection contains out-of-period films; the July 2025 agent film is excluded here. Native frame subsets establish only their stated intervals, not every boundary in a longer excerpt.
Existing Python playbook recipes establish phase and geometry contracts, not Resolve behavior. Recent Remotion research supplies useful source observations; neither its renders nor Python renders count as Resolve evidence. Existing opening, motion, overlay and ending chapters remain intact.
Original media, not borrowed product skin
Separate layers. Replaceable assets.
All media is original neutral artwork. The kit includes 12 full-frame PNG layers, reproducible asset code, redistributable DejaVu fonts and licenses. No source logos, copied demo data, private code or supplied recordings enter the reusable kit.
Download honestly
A preparation kit, not a project pack.
Download preparation kit · 0.93 MB
SHA-256 b3f8758cd8218081546891975e0a5a15089682f0e6df8b859d2a791a5aa7f88d
Agent guide (Markdown) · All recipes (Markdown) · Recipes (JSON) · Acceptance status
| Deliverable | Status |
|---|---|
| preparation kit | available; original stills, text instructions and offline helpers only |
| drp | not produced |
| self contained resolve archive | not produced |
| reusable fusion settings | not exported or imported |
| resolve recipe renders | not produced |
| resolve 52s demonstration | not produced |
| clean library restore | not run |
| rendered fidelity comparison | not run |
| fresh context adaptation | not run |
To finish: provide an authorized interactive GPU workstation with Resolve, its exact About version/edition/OS, a disposable library and export/render access. No purchase, production-library migration, unattended render or external-script permission is assumed.
Keep the two-video workflow intact.
The presentation starts first. Its recognizable content transition cues a manual switch to OBS; measure the reaction/startup delay rather than assuming instantaneous synchronization. The talking head remains separate and supplies the narration. Compare both start/end times on one timeline so the presentation ends during the genuine recorded wave and pause.
No actual cue, talking-head offset or ending-buffer length is set by this unpaired study. The guide gives the equations, marker conventions, recording-derived transcript workflow and single storyboard approval boundary. This research does not authorize a final all-hands render.