Reference F · Glossary
Reference
Every term of art a director meets in this book, in plain words: the craft's own words, the models' and the platform's. Each entry gives the meaning as the chapter that teaches it uses the word, and says which section to read.
In this reference
292 terms in alphabetical order, in six groups, each with a short meaning and the section that teaches it
The tools and models by name, and the words that name the platform's records, prices and rights
Cross-references for the words that share an entry, and for the words that carry two meanings
How to read an entry. The meaning is the one the teaching chapter gives, in a sentence or two; the chapter gives the reasoning. The last column names the section that teaches the term: 10.2 is the second section of Chapter 10. Where the chapters' own term tables define a word, the entry repeats their words. A word that means two things says so, and each meaning points to its own section. A row that reads "See" sends you to the entry that carries the word. Tools change; for a tool's current version, prices and limits read its own chapter, and the open questions are in Reference G.
F.1 A to C
Term | Meaning | See |
Alpha | The channel of an image file that carries transparency. GPT Image 2.5 is the only stills route with a real one; look at it over black, over white and over a saturated colour. | 12.7; 12.1 |
Alternate | A duplicate of a cut with one thing changed. One alternate carries one proposition. | 29.1; 29.7 |
Ambience | See Ambience, bed. | 25.1; 25.11 |
Ambience, bed | Continuous sound that holds a scene together across its cuts. | 25.1; 25.11 |
Anchor | Any approved picture a later job is built on and checked against: a master, an in-look anchor, a place plate. A drifting frame is re-anchored to the first-generation masters and never edited in place. | 10.2; 8.8 |
Anchored prompt | A still prompt for a frame built from references. It runs in order: the manifest, the identity clause, the change contract when one is needed, then the frame. | 10.5 |
Animatic | A board, or a set of stills, cut to time with sound, to judge a film before anything is generated. On an AI production the cheapest version is a stillomatic. | 4.5 |
Anti-AI watch | A field of the shot record that lists two to four specific risks of generation for the frame (a face cloning across a crowd, a rivet that looks painted on). It is a list of things to look for, and it is never pasted into the prompt. | 8.6 |
Aspect | The shape of the frame, width to height: 16:9, 9:16, 1:1. It is a setting of the job, decided first and set every time, and never a word in the prompt. | 14.15; 11.6 |
Assembly A | The first causal cut, on hard cuts, before any finishing. | 29.1; 29.6 |
Audio Influence | The Suno setting for how far uploaded audio steers the result. It appears only when audio is attached. | 26.6 |
Authority | The file that settles a question: who this person is, what the product looks like. Approve against it, never against the last view; when the words and the picture disagree, the authority wins. | 10.1; 10.2 |
Axis (the line) | The imaginary line between two people, or between a person and what they look at. The camera keeps to one side of it so that screen direction holds. | 5.3; 5.8 |
Baked | Flattened into pixels or into a mix. A baked element plays, but it can no longer be changed as an element of the edit. | 30.9; 31.5 |
Balance | The platform's report of the credits left on the account. Read it at the start of a session, before any quote. | 34.2 |
Batch | Several different requests sent to the platform together, so that they run at once and can be waited on as one. A single request suits one shot you will watch. | 33.3; 34.2 |
Blocking | Where bodies are, and where eyes go, from the first instant of a shot to the last. | 15.5; 5.1 |
Blocking photo | A quick photograph you take yourself of a hand, a grip or an arrangement, attached to a job to show what words cannot. | 9.9 |
Blockout | A coarse clip of simple shapes, one per person or object, with no axes, labels, trajectory lines or camera cones, given to a video model for its camera path and timing and not its look. A clay previz is one. | 17.3 |
Boundary state | Everything true at the seam of an extension, read from the file: pose, hands and props, setting, camera and its distance, light and the last sound. | 17.3 |
Bracket (audio tag) | A direction in square brackets inside the text, which the voice plays and does not speak. | 21.1; 23.3 |
Brand role | What the brand does in the film; intrinsic when the idea cannot exist without it. | 3.1; 3.4 |
Break shot, written ceiling | Two design tools for lip-sync exposure. A written ceiling is a project decision that caps how many visible-sync moments the locked cut may hold; a break shot sits between two of them, with voices or atmosphere and no visible mouth to match. | 22.3 |
Brief | The client's or agency's statement of the problem: the product, the audience, the objective and the mandatories. | 3.3 |
Build | The exact release of Premiere Pro that a project was made in. Record it for each project. | 32.1 |
Button | See First attack, button, cutdown. | 25.1; 25.4; 25.5 |
Cairene | Of Cairo. The book's Egyptian lines are written, cast and judged in Cairene Arabic. | 21.7; 15.8 |
Candidate | One generated version of a cue, kept for comparison with the others. | 25.5 |
Canon | The settled facts of the film's world, decided once and obeyed by every department. | 35.1 |
Canonical line | The script's words for a line, in the spoken dialect, with its punctuation; it never changes for the engine's sake. The approved text of a line of dialogue, the authority for every caption. | 21.1; 21.4; 30.7 |
Catchlight | The reflection of a light source in the eye. | 6.7 |
Change contract | The line of an anchored prompt that names where this frame differs from a reference and what still holds. | 10.5 |
Character core | What stays true of a character across the film. Only this scene's want, obstacle and trigger go into a prompt. | 8.6; 15.3 |
Character sheet | Several views of one character in one image, on flat grey, made as a reference. The 3-panel default is a headless body from the front, the body from behind, and a chest-up face close-up. | 10.6 |
Checksum | A fingerprint of a file's contents, taken with the probe and recorded beside it. | 30.2; 31.1 |
Client launch film | A launch film for a premium hair-repair mask, made for a real client, worked through in Chapter 36 as the book's product example. | 36.1 |
Clip (highlight) | See Roll-off, clip. | 6.1; 6.3 |
Clip, take | A clip is one generated video file; a take is one clip of a shot. | 13.1 |
Clip (video) | See Clip, take. | 13.1 |
Closing line | The last sentence of a prompt, ratified by the author: "no overlaid text, subtitles, logos, or interface graphics". | 2.7; 9.5 |
Colloquial Arabic «عامية» | The spoken dialect in which a character's line is written, cast and judged, as against Modern Standard Arabic. | 21.7; 21.4 |
Composite | To layer two pictures or clips so that one shows through the other, in post, where nothing else is redrawn. A mask is the outline that decides what shows. | 9.9; 14.13 |
Composition plan | Eleven Music's form of a cue: an ordered list of up to 30 chunks, each with its text, length, positive and negative styles and a context-adherence level. | 27.2 |
Conform | Converting a file deliberately to the sequence's frame rate or frame size. | 30.1; 30.2 |
Connector | The route by which an AI assistant reaches Higgsfield and its models. Higgsfield's web app and developer API may offer more than the connector does. | 33.1 |
Contact sheet | Six stills from a clip, at 0, 20, 40, 60, 80 and 100 per cent of its length, laid side by side. It finds problems; only the director's eye and ear pass the shot. | 33.7 |
Contact state | The frame before a touch, a reach or a handover, with the gap still visible, so that the clip has an action to perform. | 9.10 |
Continuity anchors | Three to five locked facts, listed in the shot record, that every later frame of the shot is checked against. | 8.6; 9.12 |
Continuity kit | The set of pictures that keep a film consistent: masters, character sheets, state references, location plates, product photographs. Build it from the shot table, and no fuller. | 8.8 |
Contract (the platform's) | On the platform, the truth about a model: what it accepts today, its settings and their defaults, and the roles its inputs may take. Read it before the first paid job of the day; what it leaves out is also a fact. | 33.2 |
Contrast ratio | How much brighter the lit side of a face is than the shadow side (2:1, 4:1). | 6.1; 6.3 |
Cover | A Suno operation that makes a new arrangement of a song you created. Judge it on what it kept before what it added. | 26.6 |
Coverage | The shots the cut needs. The missing-coverage list names each shot, what is missing, what stands in and what it costs, and sends the gap back to generation as a replacement brief. | 29.11 |
Credit | Higgsfield's unit of payment. On 6 Aug 2026 Higgsfield put 52 credits at $2.55, about five US cents each; credits have no cash value, and every price in the book carries its date. | 34.1 |
Cue | See Cue, motif. | 25.2; 25.3 |
Cue map | The locked cue's tempo, accents and section starts, in seconds, read from the file and laid on the timeline as markers. | 25.4 |
Cue, motif | A cue is one piece of music placed against picture; a motif is a short tune the score keeps returning to in changed forms. | 25.2; 25.3 |
Cue sheet | The form that records a cue's function, its windows, its theme, its roles, its refusals and its source and rights, before it is briefed. | 25.3; 26.5 |
Custom mode | The Suno mode with separate fields for the style, the lyrics and the exclusions. A finished cue is briefed in it. | 26.6 |
Cutdown | See First attack, button, cutdown. | 25.1; 25.4; 25.5 |
F.2 D to H
Term | Meaning | See |
Decision log | The running record of every decision locked, with its reason and date, and of every idea rejected, with the reason. | 8.1 |
Deck | The treatment or proposal as a sequence of pages shown to a client. | 4.8 |
Declared band | The length a prompt aims for, and why, written in the shot record. A prompt is finished when every element the shot needs is settled; the band is a diagnostic, not a target. | 8.6 |
Delivery specification | The client's or broadcaster's written specification for each version: aspect ratio, length, frame rate and resolution, codec and container, audio layout and loudness, captions and their languages, safe zones and file names. Obtain it in writing before the first frame. | 8.2 |
Derived frame | A frame made from an approved frame. Every later job attaches the first-generation masters, never a derived frame. | 10.2 |
Designed effect | A sound stylised on purpose: a sweep into a macro, a whoosh on a whip pan. | 25.1; 25.12 |
Device | The formal or structural move that carries the idea: a split screen, a repetition, a rule the film keeps. In camera work a device is a named compound move with one purpose, and it is one intention only when all five of its parts are written (14.9). | 3.1; 3.4 |
Dialogue ledger | One row for every spoken moment, joining the line and its approval, the voice, the picture and the contract. | 21.11 |
Diegetic | In the story's world (a voice, a door); non-diegetic: for the audience only (score, voice-over, a stylised whoosh). | 25.1; 25.2 |
DX, AMB, FLY, SFX, MX | The five layers: dialogue, ambience, Foley, designed effects, music. | 25.1; 25.2 |
Edit plan | The written cut: every placement with its source range, sequence range, track and reason. | 29.1; 29.2; 29.6 |
Edit (the model's) | Asking a model to redraw an existing frame, or an accepted clip, with one change. It is not an edit in the cutting room. | 10.13; 17.3 |
Editability table | The list, delivered with a handover, of what in the project is native and editable, what is baked with its source kept, and what did not carry. | 31.1; 31.5 |
EDL, AAF | Two older interchange formats for describing a cut to another application. Neither is shown to survive a handoff. | 31.4 |
Elements (Kling's) | Kling's way of binding a subject from several pictures of it. The connection does not expose them. | 13.4; 16.1 |
Eleven Music | ElevenLabs' music model (Music 2.5, music_v2_5). A cue is a plan of timed chunks. Its self-serve plans exclude film, TV and radio. | 27.1; 27.9 |
Eleven v4, Eleven v3 | ElevenLabs' speech models. Eleven v4 (eleven_v4) is the current one and Eleven v3 (eleven_v3) the previous generation; both stay in use, and the model is written on every take. Eleven v4 Turbo is built for real-time agents and the book does not use it. | 23.1 |
End frame | See Start frame, end frame. | 13.1; 14.3; 14.8 |
Endpoint, hold | The composition the clip lands on, and the still seconds after the action ends. | 13.1; 14.7 |
Enemy | The feeling, habit or norm the brand stands against; almost never a competitor. | 3.1; 3.4 |
Engine text | The same words as typed into the voice tool, with a bracket and any repairs; it never reaches captions. | 21.1; 21.4 |
Environment anchor | An anchor that fixes a place: the geography master (a wide that fixes the landmarks), coverage plates, light-state plates, material close-ups. | 8.8; 10.10 |
Event, continuation, release | A sound with a frame (a tap); a sound that lasts through an action (a scrape); the moment it lets go. | 25.1; 25.6 |
Exact take | One of the two speaking-shot contracts (Route A): the locked take ships, on a talking plate to which Sync Lipsync 3 fits the mouth. It is decided per project, with no default. | 22.1; 22.5; 8.2 |
Example labels | The four tags on a worked example, which say where its prompt comes from: Written for this book, From Mastorna corrected, Research exemplar and Maker's example. No example has been generated. | How to use this book, How to read an example |
Exclude Styles | Suno's separate field for what a piece must not contain. A cue's refusals go there, one to five sounds, each with a positive twin in the text. | 26.5; 26.2 |
Exclusions | The few things a prompt says must not appear, each aimed at a drift you expect: at most five on any prompt (three or fewer is the working advice on a clip), written after the positives. | 2.7 |
Exposure (lip sync) | How much a shot asks of lip-sync: the mouth's size in frame, its angle to the lens, the length of the line. The design of a speaking shot begins by lowering it. | 22.1; 22.3 |
Extend (a still) | To add canvas around a still. The routes, in order: originate the other shape from the approved frame, extend by hand in post, then a named outpaint model. | 10.14 |
Extension (of a clip) | New footage a model joins to one end of an accepted clip, forward or backward. It restates the boundary state briefly and adds only the next beat. | 13.2; 17.3 |
FCPXML | The XML format of Final Cut Pro 10 and later. It is a different format from Final Cut Pro 7 XML, the one Premiere imports; target one destination. | 31.4; 31.5 |
Fence | The words that stop a reference giving more than its one job: "take nothing of its room, light or pose". | 10.1 |
File manifest | The file (manifest.json) that lists, for every file of a handover, its path, duration, frame size, frame rate, streams, checksum, the interval the cut uses and its origin label. | 31.1; 8.9 |
Fill | See Key, fill. | 6.1; 6.2 |
Finish | How the frame looks as captured: grain, base tone, bloom, lens edge. It is written in the prompt as capture behaviour; the delivery texture is laid once, last, in post. | 6.6; 30.5 |
Finished grade | A prompt written to the full anatomy, for a frame or clip that will be delivered, signed or animated: twelve elements for a still, seven for a clip. It is the standard the book teaches. | 9.1; 2.3 |
Finishing | Everything done to picture and sound after the cut is locked: conform, repair, upscale, colour, texture, text, sound and versions. | 30.1 |
First attack | See First attack, button, cutdown. | 25.1; 25.4; 25.5 |
First attack, button, cutdown | The first note you hear, not the start of the file; a short definite final note, not a fade; a shorter version of the film, such as 15 seconds from 30. | 25.1; 25.4; 25.5 |
First-generation | See Master. | 10.2; 8.8 |
Flare, Sunburst | The two models of GPT Image 2.5: Flare is the small one, optimized for speed; Sunburst is the base model, optimized for quality. The names are models, not lighting effects. | 12.1 |
Foley | A contact that matches the picture, performed or recorded: a spoon, a step, a lid. | 25.1; 25.7 |
Frame 0 | The first frame of a returned clip. Compare it with the keyframe on every clip. | 14.16; 16.7 |
Frame count, frame rate | A file's number of frames and the frames it shows per second. Together they give its true length; probe both and never quote the label. | 14.15; 29.2 |
Frame inventory | The element of a still prompt that names the foreground, midground and background and what touches each edge, so that no layer is left for the model to invent. | 9.2 |
Frame study | A generated frame shown in a treatment to illustrate a beat; not casting, not a final frame. | 4.1; 4.8 |
Frozen instant | The moment of an action that a still captures. It is taken 20 to 40 per cent in, written as positions and participles, with directional evidence for every moving system. | 9.4 |
Gate | One of the four decisions that stop a production: LOCK, SPEND, RIGHTS and ACCEPT. Nothing accepts itself, and approval at one gate is not approval at another. | 8.1 |
GPT Image 2.5 | OpenAI's image model, in two models, Flare and Sunburst, reached through Higgsfield. It is the only stills route with a real alpha channel, and takes layouts and bounded edits of frames with no signed face. | 12.1 |
Grade | Two meanings. The colour correction applied to the picture (30.4). And the level a prompt is written to: the finished grade or the test grade (2.3). | 30.4; 2.3 |
Halation | The soft glow around bright highlights that a film base scatters. A prompt may describe it as capture behaviour; the delivery halation is added once, last, in post. | 30.1; 6.6; 30.5 |
Handle | Usable frames outside the interval, which a trim or a dissolve can borrow. | 29.1; 29.4; 29.11 |
Handover | What the edit ends in when it must survive beyond your own session: the native project, the collected media, stems, title sources and a reference render, with an interchange file only for a destination that has been tested. | 31.1 |
Hard, soft | The quality of a source: a small or distant one makes sharp-edged shadows, a large or close one gradual ones. | 6.1; 6.2 |
Hard sync point | A sound that must be seen and heard on the same frame. | 25.1; 25.6 |
Higgsfield | The host platform the book's stills, clips and lip-sync passes run through: one account, one credit balance, one job history and one ledger. It makes none of the book's models. | 33.1 |
Hold | The frames a model adds after the action has finished. In a speaking shot the hold is the frames after the last word, where the face stays still; the edit trims it. The still seconds after the action ends belong to the endpoint (see Endpoint, hold). | 29.1; 14.7 |
Host platform | A service that runs many makers' models under one account; the book's is Higgsfield (Chapter 33). | 1.1; 33.1 |
F.3 I to N
Term | Meaning | See |
Identity clause | Six to ten fixed facts that describe a face and its current wardrobe, asymmetries first, written once from the approved master and pasted unchanged. The master wins any disagreement. | 10.4 |
Identity stack | The set of masters a recurring character needs, made in a fixed order: the master face first, then what the shots need. | 10.3 |
In-look anchor | The character approved in the film's own light, made from the master with the master attached. | 10.2; 8.8 |
In-world text, editorial text | In-world text exists in the film's world (a shop sign, a label) and is specified and read glyph by glyph. Editorial text (a title, a caption, a legal line) is set in post from approved copy. | 10.12 |
Input interpretation | How a file's colour is read: its transfer curve, primaries and range, taken from its tags and confirmed by eye. | 30.1; 30.4 |
Instant clone, Professional clone | ElevenLabs' two kinds of clone of a real person. An Instant clone needs one to two minutes of clean speech; a Professional clone needs at least 30 minutes and is of your own voice only. Only ever of a person who signed. | 23.9; 21.6 |
Intention | The first field of the shot record: the story intent, the emotional beat and the goal of the shot. It lives in the record and is not pasted into a prompt. | 8.6 |
Interchange file | A timeline written in a format other applications can read. It describes a cut; it is not a project. | 31.1; 29.14 |
Interval | The part of a take the cut uses, written as frames: in (included) and out (excluded), counted from 0. | 29.1; 29.4 |
Invariant | What must survive the shot unchanged. | 13.1; 13.2 |
J-cut, L-cut | A cut where the incoming sound leads its picture (J) or the outgoing sound carries past its picture (L). | 29.1; 29.7 |
Job | One request to a model; each returns a result. Each accepted job has a job number. | 1.1; 33.3 |
Job record | The platform's own record of a finished job: its prompt, its settings and its model label. It is not the run record, which the operator keeps. | 33.7; 8.7 |
Join | The cut between two shots, looked at as a pair. | 29.1; 29.8 |
Kelvin (K) | The number that names a light's colour: low is warm and orange, high is cool and blue. Always a cue for the look, written beside its visible result. | 6.1; 6.4 |
Key | See Key, fill. | 6.1; 6.2 |
Key, fill | The main light on a face; the light that lifts its shadow side, here always given a cause. | 6.1; 6.2 |
Key visual | A picture that carries an idea for a treatment, a deck or a poster, with room left for type. | 9.11 |
Keyframe | A still made to be animated by a video model; not a keyframe on an edit timeline. | 1.1; 9.10 |
Kling 3.0 | Kuaishou's video model, and the camera to reach for when the move is the shot: a signed start frame, one action, one clip. It is also the usual maker of the wordless plate that the exact take needs. | 16.1 |
Kling 4.0 | Announced by Kling on 28 Sep 2026, to launch officially in October; only a limited early-access group has its first member, Kling 4.0 Flash. It is not available to use through the connector, and the book teaches Kling 3.0. | 16.1; 16.10 |
Ledger (the platform's) | The platform's list of charges. It is the only proof of what you paid. | 34.1; 34.3 |
Library voice | A ready-made voice from a voice library. It is for casting and screening; the voice in a client final is a person who signed, and whether a library voice may be used in an advertisement is not settled. | 21.3; 21.6 |
Lip sync | Making a face's mouth match a recorded audio. Sync Lipsync 3 does it as a pass on a finished video. | 22.1; 24.1 |
Lock | A decision fixed before production; a hard lock does not move, a semi-lock moves within limits. Of a take, see Take; of the picture, see Picture lock. | 4.1; 4.6; 8.1 |
Log | The record of a take's events at the frames where they happen, and of the frame where any defect starts. | 29.1; 29.4 |
Look | The whole visual character of a film: its light, colour and finish together. | 6.8 |
Lookbook | The visual references for a film's look, used by the crew from prep to shoot. | 4.1; 4.7 |
Loudness, true peak | Loudness is measured over the whole mix against the delivery owner's specification: EBU R 128 asks for −23 LUFS with a true peak not above −1 dBTP, and ATSC A/85 for −24 LKFS with a true peak not above −2 dBTP. It is never set in a generator. | 30.6; 25.15 |
LUFS | See Loudness, true peak. | 30.6; 25.15 |
LUT | A look-up table: a preset colour transform. | 30.1; 30.4 |
Lyrics field | Suno's box for the running order of a piece and any words to sing. Section tags and [Instrumental] sit in it. | 26.2; 26.3 |
Manifest | The opening sentences of an anchored prompt: one for each attached image, in order, saying what it is, what to take from it and what to take nothing of. It is written from a row of the reference manifest and is not the same thing. | 10.5 |
Mask | A black-and-white image the size of the frame that marks the only area a model may repaint, or the outline that decides what shows in a composite. Nano Banana Pro has none, so the fence is written in words and the change composited in post. | 11.6; 14.13 |
Master | The first, neutral picture of a character, a place or a product: a photograph of the actor, or a generated face on a plain grey ground in even light. It is the authority on identity, and it is first-generation: made directly for its job, never derived from another output. Every later job references it, never a frame made from it. For a place it is the authority on its geography. | 10.2; 8.8 |
Mastorna | The author's finished work and the book's worked example: a trailer for a film that was never made, reconstructed as if it had played in Italian cinemas in 1966, in black and white. | 35.1 |
Measured move | A slow camera move sized with one approximate distance or angle, a duration and a speed profile, with the visible result beside it. It is written on Kling and measured on the returned clip. | 14.5 |
Mechanism | What the audience does while watching: laughs at a repetition, waits for a reveal, does the arithmetic before the film does. In camera work the mechanism is the body that gives a move its weight: a dolly, a slider, a crane arm, named for its behaviour and never its brand (5.4). | 3.1; 3.4 |
Mid-action status | The field of the shot record that lists the frozen systems of a moving shot and how far into its action each one is, or STATIC with the reason. It is the one field the prompt must repeat. | 8.6 |
Mode | The setting that selects the operation on a model. The route line records the model and the mode together. | 13.3; 17.3 |
Model | The program that makes the picture, clip, voice or music (Nano Banana Pro, Kling 3.0, Eleven v4 …). | 1.1; 2.1; 33.1 |
MOGRT | A Premiere motion graphics template: an editable graphic whose text can be retyped. | 30.1; 30.9 |
Motivated | See Motivated, practical. | 6.1; 6.2 |
Motivated, practical | Lit by a source the scene can explain; a light that is part of the set (a bulb, a lamp, a screen). | 6.1; 6.2 |
Move | Anything the camera does between the first frame and the last. It is written as a verb, a mechanism, a destination and a hold. | 5.4 |
MSA (Modern Standard Arabic) «الفصحى» | The formal written language, as against the colloquial dialect of a character. A line of dialogue for a character is written in the dialect. | 21.7; 21.4 |
Music-first, picture-first | Which leads: the cut is made to a locked cue, or the cue is made to a locked cut. The edit owns all sound in both. | 25.4 |
Najdi, Hejazi | The Arabic of Riyadh and of Jeddah, two Saudi varieties, cast separately. | 21.1; 21.8 |
Nano Banana Pro | Google's image model (Gemini 3 Pro Image), run on Higgsfield: the stills camera of the book. It makes the hero frames, the masters and anchors, and the keyframes; it has no mask and no seed. | 11.1 |
Native audio, own sound | The soundtrack a video model generates with the picture. The edit owns the final sound, so set the sound switch on every job and treat what the model returns as scratch. | 25.1; 16.6 |
Native ear, dialect reviewer | A listener who speaks the dialect natively. For every new Arabic voice the native ear is the test, and the dialect reviewer approves every mark on a word before it runs. | 21.6; 21.7 |
Native original | The file as generated and accepted, kept untouched beside any transformed copy. | 30.1 |
Native project | The project file Premiere itself saves: bins, sequences, effects and links to media. It is the first thing a handover delivers. | 31.1 |
New performance | The other speaking-shot contract (Route B): a video model performs the line from the signed frame and the voice, and the audio it returns ships once the director has judged it on the clip. | 22.1; 22.6; 8.2 |
Not yet known | No source answers it. The chapter says what to do meanwhile, and Reference G collects the cases. | How to use this book, Dates, prices and doubt |
F.4 O to R
Term | Meaning | See |
Observation | A fact about behaviour with no tension in it ("Gen Z uses TikTok"): true, and inert. | 3.4 |
Omni Edit | Kling's video-edit model on Higgsfield (kling_video_edit): a source clip and optional picture references, for one scoped change inside a clip you have already accepted. | 16.6; 16.9 |
Onset, closure | The first visible movement of speech, and the lips meeting, as on ب and م (b and m). Both are checked frame by frame after a lip-sync pass. | 22.7; 24.7 |
Operation | What a job does: make a shot from words, animate a frame, travel between two frames, compose from references, edit a clip, extend a clip. | 13.1; 13.2 |
Operator | Whoever writes and runs the jobs: the director, an AI artist, or an AI assistant working for the director. | 1.1; 1.5 |
Origin label | The record of where a frame or clip came from: generated, upscaled, painted over, composited, practical, client-supplied or stock. With the checksum it makes every clip traceable to its source and route. | 31.1; 8.7 |
Originate | To make a frame again at a new shape, as a new composition with the approved frame attached, instead of extending its canvas. | 10.14 |
Palette plan | A split such as 60/30/10: about 60 per cent of the frame in a dominant colour, 30 in a secondary, 10 in an accent. It is planned as proportions and written in the prompt as colours on named surfaces, never as the string "60/30/10". | 6.5; 8.6 |
Parallax | Near things passing faster than far ones as the camera travels. | 5.5 |
Parent | The frame a job started from, and the source of truth for everything the job does not name. In Suno every operation makes a child version from an approved parent. | 10.13; 8.7; 26.6 |
Pass | One Sync Lipsync 3 job: a video and a take go in; the same video comes back with a new mouth. | 22.1; 24.1 |
Person who signed | The rule for a voice in a client final: a consenting actor, or a clone of that person under a written release. Never a library voice, and never a clone of a voice from a public clip. | 21.6; 34.7 |
Persona (Suno) | A saved vocal and style character of an existing song, filed with Suno's voices. Never save one made from client material. | 26.6; 26.9 |
Perspective (sound) | How far away a sound seems and in what room, including off-screen sound. It is written into the effect's description. | 25.6; 25.7 |
Phase | How far into its action each moving part is, written as positions ("rear foot just released"). In the Mastorna chapter a phase is also one of twelve steps of the author's brief; the work stops at the end of each until the author says to continue. | 9.4; 35.1 |
Picture lock | The point at which the picture is agreed and every later step works on it without changing it. | 29.1; 29.14 |
Plate | A still of a room or a product, with no action, made to be reused as a reference. In motion the word also names a clip made for a later pass: a talking plate for lip sync, a wordless plate (Chapters 13 and 22). | 10.2; 10.10; 13.2; 22.4 |
Post-dubbed | Speech recorded afterwards and laid to the picture, as Italian cinema of 1966 did. | 35.8 |
Practical | See Motivated, practical. | 6.1; 6.2 |
Premiere Pro | The editing program the cut is built, judged, finished and exported in. The assistant plans and checks; Premiere executes and confirms. | 32.1; 29.1 |
Preserve list | What stays in an edit, named region by region: faces, hands, product, set, camera, crop, light, grain. It is repeated on every edit. | 10.13 |
Preset | A ready-made recipe that the platform may offer in place of running your request. A preset suggestion is not a job: your request has not run. | 33.2 |
Previz | Previsualisation: a rough animation of a shot, such as grey clay shapes, used to plan it or to give a video model a camera path. | 17.3 |
Probe | Measuring a file with a tool such as ffprobe: size, frame rate, frame count, length, audio streams. | 29.1; 29.2; 14.15 |
Project decisions card | The one card on which the decisions made at the start of every project are written and locked at the first LOCK gate: the speaking-shot contract, the delivery specification, the rights class and the look. | 8.2 |
Prompt | The written brief a model reads for one job. | 1.1; 2.1 |
Pronunciation dictionary (alias) | An ElevenLabs list that swaps a written word for a spelling the voice reads correctly. It suits a brand name or a number that recurs. | 23.3; 23.2 |
Protected moment | A beat of the approved treatment that no later version may change. | 4.1; 4.6 |
Provider | The catalogue field that says who made a model or tool. A tool whose engine is not named is not used. | 33.2 |
Quiet zone | The part of a frame kept calm for type, written as a plain surface. | 9.11; 12.5 |
Quote | A price check on an exact request, made at no cost with its pictures attached, that starts no job. Plan from the dated table, spend from today's quote, settle from the ledger. | 34.1; 34.2 |
Raster | The pixel size a file actually has, width by height, as measured. A resolution label names a size; the raster is the fact. | 14.15; 30.2 |
Re-roll | Running a prompt again for a fresh result. The retry ladder allows one, unchanged, because variance is real. | 14.16 |
Reference | A picture, clip or recording you attach so the model can borrow from it: a face, a room, a product, a voice. It may lend a face, a product, a place, a camera path or a voice without fixing the frame. | 1.1; 13.1; 10.1 |
Reference manifest | The single list every later prompt and every check reads: one row for every recurring picture of a character, place or product, with its identifier, version, status, role, subject label and route. | 8.8 |
Reference render | A render of the delivered project, handed over so that the next editor can see what the timeline should produce. | 31.1 |
Reframe | To recompose a shot for another aspect ratio. Reframe by hand before trying automatic tracking; a shot that cannot be cropped is originated again. | 30.8 |
Refusal | Something the category always does that this idea will not. | 3.1; 3.5 |
Relight | The same frame under a new light state, made as an edit from the approved frame. | 10.14 |
Relink | Reconnecting a project to its media after the files have moved. | 8.9; 31.5 |
Replace Section, Song Editor, Studio | Suno's repair surfaces. Replace Section returns alternatives for a selected region; the Song Editor adds, removes, rearranges and fades sections; Studio is the multitrack editing surface. Repair the smallest region that fixes the fault before you regenerate. | 26.6 |
Resolution (the setting) | A label such as 480p, 720p or Kling's std and pro. It names a size and does not guarantee one; probe the file. | 14.15; 16.6 |
Retry ladder | The order of repairs for a failed take, smallest lever first: re-roll once unchanged, reduce the intensity, strengthen the protection, simplify to one motion system, re-phase, split the shot, return to the keyframe. The same failure twice means change the input, not the adjectives. | 14.16 |
Reverse angle | The camera turned to face the other way across a scene, usually onto the other person. On a reverse, world positions stay and screen positions change. | 5.3; 5.8 |
Revision levers | A field of the shot record: two or three edits planned before the first attempt, so that a near miss gets a named lever and not a regeneration. | 8.6 |
Rig | The machine a camera move is imagined on: a dolly on boards, a crane arm, a hand-operated head. It is named for its behaviour, never its brand, and named first. | 14.4; 5.4 |
Rights class | What a job will contain, classified before anything is uploaded: neutral synthetic, private or rights-cleared, client-confidential, real person or biometric, voice, brand or trademark, licensed third party, public domain, or prohibited or unverified. | 8.2; 8.10 |
Rip reel | A sequence cut from existing films to show pace and tone; internal only. | 4.1; 4.7 |
Roll-off | See Roll-off, clip. | 6.1; 6.3 |
Roll-off, clip | How light fades across a curved surface such as a cheek; a highlight so bright it keeps no detail. | 6.1; 6.3 |
Room tone | The sound of the room with nothing happening in it: the floor under silences and dialogue edits. | 25.1; 25.11 |
Route | A way of making something: a model and a method ("a start frame on Kling, its sound set"). | 1.1; 13.6 |
Route A, Route B | The two ways to make a speaking shot. Route A is the exact take: the locked take ships, on a talking plate. Route B is a new performance: a video model performs the line. | 22.1 |
Route line | The one line, kept in the shot table and the shot record and carried into the run record, that records the choice of operation. It has eight parts: invariant, operation, grade, model and mode, inputs in order, settings, reason and fallback. | 13.3 |
Run record | The log of every attempt, failures included: the request as sent, the settings, the inputs, the job number, the charge, the file, its hash and the verdict. It is written the day each attempt runs. | 8.7 |
Rung | One step of the repair ladder (21.7); a line climbs only as far as the failing word requires. | 21.1; 21.7 |
Running rough cut | The cut that grows as takes are accepted, where every new take is judged in context. | 1.1; 29.10 |
F.5 S
Term | Meaning | See |
Safe zone | The part of a frame that stays clear of a platform's interface. No platform figure is given in the book: ask the delivery owner for the overlay and measure your own frames against it. | 30.8 |
Screen direction | The side a subject enters from and leaves by, planned so that neighbouring shots cut together. | 14.12; 29.8 |
Seedance 2.5 | ByteDance's video model, which makes picture and sound in one pass. It takes pictures, clips and sound as references, composes a shot from anchors with no start frame, and can speak an English line in a chosen voice. | 17.1 |
Select | An interval chosen for the cut. A changed select, a different take or a different interval of the same take, costs nothing and comes before a new clip. | 29.11 |
Sheet | Several views of one character, place or object in one image. Attach a sheet only where the route reads it; otherwise crop each view to its own file. | 10.6 |
Shot record | The director's page for one shot: its intention, parameters, reference strategy, continuity anchors, source-lock check, anti-AI watch, mid-action status, revision levers and declared band, with the exact prompt. It is never pasted into the prompt whole. It is the second of the three linked records. | 8.6 |
Shot table | The plan: one row for every shot the film needs, short enough to read on one screen. It is the first of the three linked records, and each row holds the shot's route line. | 8.5 |
Sign | Approve a still, a take or a cut as final; everything later is built from it and checked against it. | 1.1; 9.12 |
Slate | The label an export carries, saying what version it is and what decision it asks for. "DRAFT — NOT FOR APPROVAL" marks work in progress; "ROUGH CUT vN — FOR EDIT DECISION" names the question. | 31.2 |
Slug | A placeholder in the timeline marking coverage that does not exist yet. | 29.1; 29.6 |
Soft | See Hard, soft. | 6.1; 6.2 |
Sound contract | The written decision, made before a shot's prompt, about whose sound the shot carries: a silent picture, the model's own sound, the exact take or a new performance. | 21.2 |
Sound spine | A film's plan for its sound, written before the cutting: motifs, the five layers, and silence as a cue. | 25.2 |
Source | The thing in the world that makes a light: a bulb, a window, the sun, headlights, a screen. | 6.1; 6.2 |
Source class | How a voice came to be: actor recording, consenting clone, library, designed or converted. | 21.1; 21.3; 21.6 |
Source-lock check | A field of the shot record: the must-show items the frame has to contain, ticked from the source-lock packet. | 8.6 |
Source-lock packet | For one scene, what the source says must be seen and what must not drift. | 35.4 |
Speaking frame | A keyframe designed for a mouth that will be made to speak. The easiest frame to sync is not necessarily the prettiest keyframe. | 9.10 |
Speaking-shot contract | The project's written decision on how a speaking shot is made: the exact take ships, or a new performance is acceptable. There is no default; it is decided at the start, by shot class if need be, and a change is a written route change. | 8.2; 21.2; 22.1 |
Specification (camera and light) | The camera and the light written as numbers (height, distance, lens, stop, Kelvin, one contrast ratio), each with the visible result beside it. The numbers are cues for the look, never the only carrier. | 2.5; 9.3 |
Speech window, talk window | Where the speech starts and ends in the take; where the actor's own lips move in the plate. Both are measured, and the speech should start where the talk starts. | 22.1; 22.4 |
Spending packet | The page that opens every paid batch and is signed before anything is spent: its purpose, route, rights class, live price, cap, retry allowance, stop conditions and approver. | 8.10 |
Spotting | Listing every sound event on the selected take, with its frame and its owner, before any sound is sourced. | 25.6 |
Stable Audio 3 | Stability AI's audio model. It makes an effect from words, makes a new version of a recording you made, or remakes one region. It is not a route until the director opens an account. | 28.3; 28.1 |
Start frame | See Start frame, end frame. | 13.1; 14.3; 14.8 |
Start frame, end frame | A signed still the clip must begin on, or land on. | 13.1; 14.3; 14.8 |
State 0 | The instant before the action: the frame a clip starts from. | 13.1; 14.3 |
State (character) | A named version of a character or object that the story needs: wet after rain, a jacket taken off, a seal broken. Each state is made as a variant of the clean master, never of the last shot. | 10.7 |
State window | The span of the story in which a look state (wet, torn, in the coat) is true; it changes only at a stated beat, and each state has its own master. | 8.8 |
std, pro, 4k | Kling's working modes on the connector. The 4k mode is offered and has not been probed; quote the probed raster, never the label. | 16.6 |
Stem | One part of a piece separated from the finished mix. Suno's stems are separations of the mix and not the original tracks, so they will not add back to exactly the full mix. A mix stem is one separated layer of the final mix. | 26.6; 25.16 |
Style field | Suno's box for what the whole piece sounds like, written as one directed paragraph in the eight parts of the brief. | 26.3 |
Sunburst | See Flare, Sunburst. | 12.1 |
Suno | A music generator, run on suno.com on your own paid account and not on Higgsfield. A cue is briefed in Custom mode like a composer's brief, with no artist or composer named. | 26.1 |
Sync Lipsync 3 (Sync-3) | sync.so's lip-sync model, hosted on Higgsfield. A finished video and an audio file go in; the same video comes back with the mouth, and the face around it, redrawn to the audio. It has no prompt, so all the direction goes into the plate. | 24.1 |
sync_mode | The Sync Lipsync 3 setting that decides what happens when the plate and the take differ in length. The book sets it to cut_off on every pass and never leaves the default. | 24.2; 24.6 |
F.6 T to W
Term | Meaning | See |
Take | One generated or recorded reading of a line; lock is the accepted take, normalised, named and measured. In the lip-sync chapters the take is the locked recording of the line, prepared for the pass. In the edit, a take is one clip of a shot (see Clip, take). | 21.1; 21.7; 29.4 |
Take ledger | One row for every generated take of a line: the model, every setting, the failing word and the rung that repaired it, the file and the verdict. | 21.11 |
Talking plate | A clip of one face talking quietly with no words, made for a lip-sync pass to redraw; "plate" alone in this Part means this. It lands on a held state. | 22.1; 22.4; 24.4 |
Tashkeel «تشكيل» | Arabic diacritics: short-vowel marks, sukūn (no vowel) and shadda (a doubled consonant). | 21.1; 21.7 |
Tech read | Anyone's technical observation of a result, usually the assistant's. It is kept in its own field and is never promoted to a verdict. | 8.7; 1.3 |
Tension | Two things pulling against each other, written as one two-sided sentence. | 3.1; 3.4 |
Test grade | A prompt that keeps only the essentials. It proves a route or a look cheaply and is never presented as the model of a finished frame or clip. | 9.6; 2.3 |
Text to Dialogue | An ElevenLabs tool that makes several voices speak an exchange in one pass. It is a rhythm pass: every line that ships is remade as its own take and locked. | 23.3 |
Text to video (t2v) | A shot made from words alone, with no start frame, so that the model designs the frame too. It gives up identity and exact composition and earns its place on four grounds. | 13.4 |
Texture | Film grain and halation, added in post. | 30.1; 30.5 |
Topaz, ByteDance (upscalers) | The two upscaling engines the platform names, beside Topaz on your own machine. A frame-rate boost is a separate setting. | 32.6; 32.1 |
Transient | See Transient, attack, decay. | 25.1; 25.7 |
Transient, attack, decay | The sharp front of a sound, the way it starts, and the way it dies. | 25.1; 25.7 |
Treatment | The director's document for one film: thesis, scene breakdown, world, method. | 4.1; 4.3 |
Trigger | The visible or audible event that starts a change in a shot, such as the light that comes on when a door opens. In a performance, the trigger is what makes a character change. | 14.9; 15.3 |
Twelve elements, seven elements | The two anatomies of a finished-grade prompt. A still settles twelve elements in a fixed order, world first, photography second, technical close; a clip settles seven, from the declaration and its rig to earned exclusions. | 9.2; 14.4 |
Upscale | Making a larger file from a smaller one; the upscaler invents the new detail. | 30.1; 30.3 |
Variant | One of the results returned by a single request. On GPT Image 2.5 the word also names which of the model's two versions runs the job (Flare or Sunburst). | 12.6; 11.6 |
Veo 3.1, MiniMax H3 | Two video models on the bench: priced and read, and roles to test. Veo 3.1 (Google) takes a sound-led English action with sound always on; MiniMax H3 offers a 2K picture from a keyframe. | 19.1 |
Verdict | The director's judgement of a result: accepted, rejected with what failed, or pending. Anyone else's reading is a tech read. | 8.7 |
Voice Changer | An ElevenLabs tool that keeps a performer's dialect and cadence and changes the voice. Name the model, eleven_multilingual_sts_v2. | 23.6; 23.2 |
Voice Design | An ElevenLabs tool that makes a voice from a written description. It has no person behind it, so it is an audition, and never a voice for a final. | 23.6; 23.3 |
Voice ID | The identifier of a voice on ElevenLabs. It is written on every take, beside the voice's source class. | 21.11; 23.4 |
Wan 3.0 | Alibaba's video model, and the economy route: a plate or a cheap probe of one camera sentence, and an English line in a chosen voice. It is never used for Arabic. | 18.1 |
Window (music) | A stretch of picture, in seconds, where something must happen (the entry, a line, a contact, the turn, the exit), marked hard, it must land, or flexible, a range. | 25.3 |
Wordless plate | A plate made with no speech, timed to a take, for the exact take to fit a mouth to. It is a talking plate when a face talks quietly in it. | 13.2; 35.8 |
Next: Reference G · Still open




Comments