Crate ffcine_breakdown 

ffcine breakdown: the shot list of a reference video (a reel, a trailer), for recreating it.

One decode at analysis size (NVDEC with its hardware resize when it takes the clip, else the CPU) feeds the scene detector of ffcine triage frame by frame, the camera motion estimate (motion), per-frame sharpness and colour (colour). From the cuts come the shots; a second decode at keyframe size writes each shot’s first, sharpest-middle and last frame. The soundtrack’s beats (ffcine_audio::beats) say whether the cuts land on them (rhythm), batched faster-whisper (as triage) transcribes the speech, and tesseract, when on PATH, reads on-screen text off the keyframes. Out come breakdown.json, contact-sheet.jpg (sheet) and shotlist.md (shotlist); see docs/breakdown.md.

Re-exports

pub use colour::Colour;

pub use motion::Motion;

pub use motion::MotionKind;

pub use motion::MotionParams;

pub use rhythm::Music;

Modules

colour

Per-frame colour samples and their per-shot average.

motion

Camera motion from small luma frames: a global shift per frame pair (coarse-to-fine SAD over a three-level pyramid), refined per textured block, and a zoom fitted to the blocks’ divergence. Per shot the frame-pair motion is summed into pan, tilt and zoom, and its high-frequency part is the handheld shake.

ocr

On-screen text through the tesseract CLI, when it is installed.

rhythm

Whether the cuts land on the music’s beats.

sheet

The contact sheet: every shot’s keyframe in a grid, labelled with its number and length.

shotlist

shotlist.md: the shots as a Markdown table with an empty notes column to fill in.

Structs

Breakdown

breakdown.json.

Keyframes

A shot’s stills, relative to the output directory.

Pacing

Shot lengths over the whole video.

Shot

One shot, between two cuts.

Span

A shot’s frame range and its chosen stills, before the analysis is attached.

Transcript

Enums

Error

Constants

VERSION

breakdown.json format version.

Functions

clock

m:ss.cc.

pacing

Shot-length statistics.

spans

Splits n frames at cuts (frame indices where a shot starts, ascending) and picks each shot’s stills: one frame in from each end (so a soft cut doesn’t leak in), and the sharpest frame of the middle 60 %, favouring the centre.

words_in

The words of t whose middle falls in [start, end), with the text of segments that have no word times, in time order.

Type Aliases

Result