Section 9: AI Video Editing — the Same Edit in DaVinci and CapCut, Ten DaVinci AI Tools, Premiere's Generative Features, SubMagic, Autopod, Adobe Enhance, Voice-Clone ADR, and Horizontal-to-Vertical Outpainting

Import, timeline, transport keys; cut with Cmd+B; project settings (resolution, frame rate). B-roll: set in and out points with I and O on the source before dragging, then align to the word it illustrates - not 'drag the whole clip and delete'; layers stack top over bottom, the top clip plays; Pexels for free stock. Punch-in every other same-speaker clip to fake a second camera, matching eyelines across the cut; DaVinci: copy, select targets, Option+V paste attributes (position/zoom/colour/audio); CapCut: per clip. Gaps: DaVinci Cmd+Shift+[ / ] ripple-deletes left or right; CapCut Split (Cmd+B), Q delete-left, W delete-right. Transitions (Push, glare, pull in/out; motion blur and ease), main-track magnet in CapCut, slow motion and rotation, before/after snap builds. Music from Epidemic Sound, volume and fades. Export/Deliver.

Colour: apply a camera-matched LUT first (drop the file in DaVinci's LUT folder, right-click Refresh), then wheels - Lift shadows, Gamma midtones and skin, Gain highlights, Offset everything; dragging a wheel also shifts white balance along orange/blue and magenta/green. DaVinci's Primaries 'A' button auto-corrects. Captions: auto in CapCut and Premiere, Studio-only in DaVinci (free fallback: manual subtitle track, type and cut); editing the Track (DaVinci) or with 'apply to all' (CapCut) restyles every caption, while Customize caption / apply-to-all unchecked styles one; style with font, size, background box and radius, stroke, shadow, position, caps; split a long caption with the Split tool and move text between halves; Premiere: ~15 characters max, one line, near-zero min duration and gap for the social look. Keyframes: the diamond next to Zoom/Position - set one value, move the playhead, set another; DaVinci's Keyframes window with the curve button eases the motion.

DaVinci: Fusion Magic Mask (Faster/Better, scribble, subtract, Track) or a pen-tool mask with Add Alpha Output and the tracker; CapCut: Video > Cutout > Auto Cutout, one toggle. Feather the edge. Uses: background replacement, picture-in-picture graphics behind a cutout, the clone effect (duplicate the cutout and reposition), a keyframed slide-off. Text intros with stagger animation, pop transitions, downloaded title templates (contentcreatortemplates.com), subscribe overlays, film-burn screen composites, glitch transitions, paper-fold and logo slide-ups, motion backgrounds, compound clips for multi-layer transitions. Sound design last: whoosh on motion (time the middle of the whoosh to the middle of the move), pop on graphics, boing and crickets for comedy, cash register for money lines; stacked SFX tracks so sounds do not cut each other off; 'sound design is definitely one of those things most beginners will forget, but it adds a huge level of professionalism.'

(1) Magic Mask cutout with tracking. (2) Transcribe - edit by text, insert or remove-silence buttons. (3) Create Subtitles from Audio with track/caption styling. (4) AI Voice Isolation (Inspector > Audio, 0-100) for wind and street noise. (5) Smart Reframe follows the point of interest for vertical. (6) Super Scale 2x/3x/4x with sharpness and noise reduction. (7) AI Music Remixer - mute drums or synths with sliders. (8) Auto colour on the Primaries wheels. (9) Object Removal: Color page mask -> Tracker forward and back -> Corrector node -> Object Removal -> Scene Analysis -> Build Clean Plate. (10) Speed Warp / Optical Flow motion estimation for clean slow motion.

Transcribe, then delete text to delete video; the transcript filter finds Pauses and filler words (um, so, okay) - Extract ripples the timeline, Lift leaves a gap. Create Captions with ~15 characters, single line, near-zero minimums. Generative Extend grows a clip at the front or back with new photoreal frames (horse hair, a drone waterfall) - 'you can't even see the transition between what's real and what's not.' Remix (under the Ripple Edit flyout) lengthens music by looping repeatable sections, shown as jagged marks on the waveform. Enhance Speech in Essential Sound > Dialogue, mix amount 1.5-3 to stay natural. Scene Edit Detection auto-cuts a pre-stitched long clip. Auto Reframe Sequence with motion tracking for vertical. The lesson is maintained as features ship.

SubMagic (short form): upload up to 5 min / 2 GB; Trim (Remove Silences 'extra fast', Remove Bad Takes) -> Caption (presets like 'Ella', emoji auto-insert, per-word edits, Clean Audio, brand, music) -> B-roll (scene breakdown, search or upload, 3-5 second rule, or Magic B-roll by coverage percentage), transitions with paired SFX, Magic Zooms, export 1080p-4K; Magic Clips pulls viral moments from a long-form URL scored out of 100 with multicam speaker switching; Opus Clip and Captions are the rivals. Autopod (autopod.fm, ~$29/mo, Premiere-only): name speaker tracks (A1/A2, V1-V3), set cutting method, wide-shot frequency, minimum cut length; every camera must be synced (Premiere right-click Synchronize); Create Multicam Edit - light blue enabled, dark blue disabled. Adobe Podcast Enhance Speech V2: free tier audio-only, 30 min, 500 MB, 1 h/day (Pro: video, unlimited, in Creative Cloud); export audio, enhance, download, drop over the muted original; do it at the END of the edit so re-cuts do not reprocess; CapCut's equivalent moved to Pro; DaVinci's Voice Isolation targets noise rather than dialogue.

ElevenLabs Voices > Add New Voice > Professional Voice Clone (Creator plan; one slot). Training audio must match the target's microphone and energy - 'they look for the style and the way that we talk.' Source 30+ minutes from existing footage: remove pauses, loop short footage to reach the floor, export audio-only; 30 min minimum, 1 h better, 2 h best; record the mic-verification line; wait 2-6 hours. Then Text to Speech with Speed, Stability (lower = more expressive), Similarity, Style Exaggeration; a short line cost 49 credits with two free regenerations; drop it in the timeline over the flubbed word.

Works on a locked tripod shot with clean margins. Export one frame; in Photoshop flip the crop 16:9 to 9:16, Generative Expand (with or without a prompt), pick the best of three, spot-fix with Generative Fill 'remove'; expanding section by section (bottom, then top) beats one pass; export as sRGB. In the NLE swap the sequence's resolution numbers, place the expanded still under the original clip, mask and feather around the subject with Opacity, nest as a compound clip. Free alternative: Pixelcut Reframe to 9:16, regenerate until it looks right (email-gated download).