AI Filmmaking Glossary: 60 Terms Every Brand Marketer Should Know

A working vocabulary for brand marketers, producers, and creative directors evaluating, briefing, or commissioning AI video production work. The terms a senior production studio uses in conversation, defined for the marketing-side reader who is encountering them for the first time.

This is a reference post — bookmark, search, return when needed. Terms are organized by category rather than alphabetized, because grouping by concept is more useful than grouping by spelling.

Generation models and tooling

AI video model: A foundational model that generates video from text prompts and/or image references. Examples: Veo 3, Kling 2.0, Runway Gen-4, Sora 2, Hailuo, Seedance.

Text-to-video (T2V): Generation method where the input is purely a text description and the output is video. The simplest workflow but offers the least control.

Image-to-video (I2V): Generation where the input is a reference image (often a still frame the studio wants to animate) plus a text description of the motion. Provides much stronger control over visual consistency.

Video-to-video (V2V): Generation that takes input video footage and modifies it stylistically while preserving motion. Used for style transfer, animation effects, and aesthetic transformation of live-action plates.

Reference image / reference frame: A still image fed into the generation as a visual anchor. Used to lock character likeness, environment, product details, or aesthetic. Critical for multi-shot consistency.

LoRA (Low-Rank Adaptation): A small, trainable model layer added to a base model to teach it specific characters, styles, or aesthetics. Used for brand mascot consistency or director-specific look development.

Multi-model pipeline: A production workflow that uses multiple AI video models in sequence or per-shot, casting each model to the task it does best. Standard for premium brand cinema production.

Premium tier model: AI video models that operate at the highest current quality standard. As of 2026, this is Veo 3, Kling 2.0, Runway Gen-4, and Sora 2.

Prompting and direction

Prompt: The text description fed into an AI video model to produce output. The instruction the model follows.

Prompt engineering: The craft of writing prompts that produce intended output. Involves vocabulary choice, structural ordering, balance of specificity and ambiguity, and model-specific syntax.

Prompt template: A reusable prompt structure with variables that can be filled per shot. Used to produce consistent style across many shots in the same project.

Negative prompt: An explicit specification of what should not appear in the generation. Used to exclude common failure modes (e.g., "no extra fingers, no melting faces").

Style reference: A reference image used to guide the aesthetic of the generation, separate from a content reference. Often a still from a film, a painting, or a photograph.

Director-style reference: A specific filmmaker or visual artist whose work guides the generation. "In the style of Wong Kar-wai," "shot like a Roger Deakins frame."

Shot structure and composition

Shot list: A document specifying every distinct shot in a project, with composition, intent, length, and model assignment per shot. Standard pre-production deliverable.

Hero shot: The most important shot in a project. The frame that anchors the brand identity, lands the message, or carries the emotional weight. Hero shots get more generation attention and budget than supporting shots.

Establishing shot: The opening shot of a scene that orients the audience to the world. In AI production, often generated separately from action shots and used as visual reference for subsequent shots in the same world.

B-roll: Supporting footage that supplements primary shots. In AI production, often generated as variant shots of the same world, used as cutaways or transitional material.

Pickup shot: A shot generated late in production to fill a gap, replace a failed shot, or address a revision. In AI production, faster and cheaper than in traditional production.

Coverage: Multiple shots covering the same scene from different angles. AI production handles coverage via multi-shot generation around a consistent world reference.

Character and continuity

Character locking: The technical workflow of maintaining visual consistency of a recurring character across multiple shots. Uses reference images, LoRA training, or platform-specific character library tools.

Continuity drift: The gradual change in visual details (character likeness, environment, color, lighting) across a sequence of generated shots. The primary technical problem of multi-shot AI production.

Reference engineering: The pre-production work of preparing reference images that anchor character, environment, and style for downstream generation. A senior craft skill in modern AI production.

Cross-shot consistency: The state of having visual continuity (character, environment, color, lighting) maintained across all shots in a sequence. The mark of competent AI production.

Talent likeness work: Production work involving the visual representation of a specific real person (founder, celebrity, executive). Requires reference photos, licensing, brand-safety review, and senior compositing.

Production phases

Pre-production: All work before AI generation begins — brief, treatment, storyboard, shot list, references, model assignments. Where the project is won or lost.

Generation phase: The phase of producing AI video output across the shot list. Typically 1-3 weeks for a commercial brief.

Curation: The work of reviewing AI generation candidates and selecting the best output per shot. Done by senior creative direction, not the studio's junior staff.

Vision-grade scoring: A workflow where AI vision models (e.g., Claude Vision, GPT-4 Vision) score generation candidates against brief criteria. Used for first-pass curation before human review.

Re-roll: Re-generating a shot that failed to meet brief standards on first generation. AI generation has typical failure rates of 30-50% for stylized work, requiring multiple re-rolls per finished shot.

Compositing: The post-production work of integrating AI-generated elements with each other and with any live-action footage. Senior craft work using tools like After Effects, Nuke, or DaVinci Resolve.

Color grade / Color: The cinematic color treatment that pulls multi-shot output into a consistent visual world. The single most important phase for distinguishing premium AI production from generic AI output.

Sound design: The creation of all sonic elements in the work — ambient, foley, designed effects, transitions. Distinct from music.

Score / Music: The musical layer of the work. Either licensed (existing tracks paid for use rights) or custom (composed for the project).

Mix / Mastering: The final balancing of all audio elements. Standard finish process.

Master deliverable: The final approved version of the work in highest-quality format. All variants derive from the master.

Aspect ratio variant: A version of the work re-framed for a different aspect ratio (16:9, 9:16, 1:1, 4:5, 21:9 cinema). Each variant typically requires re-framing or re-generation, adding 5-15% to project cost.

Quality and craft

Brief fidelity: The degree to which generated output matches the original brief. The primary quality metric for commissioned work.

Brand-safety: The state of generated content not containing problematic elements — competitor marks, real-person likenesses without license, regional cultural sensitivities, brand color drift. A required production phase for commercial work.

Photorealism: Generation quality where output is indistinguishable from real photography or cinematography. The current capability frontier in AI video; Veo 3 leads this dimension as of 2026.

Aesthetic control: The degree to which the output matches a specific intended visual style, beyond just photorealism. Kling 2.0 leads this dimension as of 2026.

Cinematic quality: A subjective measure combining photorealism, aesthetic control, camera language, color, and sound. The standard premium brand work aspires to.

AI artifact: Visual flaws introduced by AI generation — extra fingers, melting features, environment pops, motion glitches, text drift. Cleanup of artifacts is a meaningful phase of post-production.

Hallucinated detail: An element generated by AI that was not requested or that misrepresents real-world fact (a logo with wrong shape, a clock showing wrong time, a city skyline with incorrect landmarks). A specific failure mode of AI production.

Audio and dialogue

Native audio sync: The capability of an AI video model to generate video and matching audio (including dialogue with lip-sync) in a single pass. Veo 3's defining capability as of 2026.

Voice cloning / Voice generation: The creation of synthetic speech in a specific voice, used for narration, dialogue, or talent-likeness audio.

ADR (Automatic Dialogue Replacement): The traditional film term for replacing dialogue audio in post-production. Sometimes still used in AI production for replacing model-generated dialogue with higher-quality recorded dialogue.

Foley: The creation of sound effects (footsteps, fabric, prop sounds) recorded or designed in post-production. Required even when AI generates ambient and dialogue audio.

Business and engagement terms

RFP (Request for Proposal): A formal document brands send to multiple studios soliciting proposals for a project. Standard procurement workflow for projects above $50,000.

Studio rate card: A studio's published pricing across project tiers. Mature studios share rate cards openly; less mature studios negotiate every project from scratch.

Pre-production fee: A separate fee for the brief, treatment, and shot-list phase, sometimes charged before the generation work begins. Increasingly common for projects above $30,000.

Revision round: A formal opportunity for the brand to provide feedback and request changes. Two revision rounds is standard for studio engagement; one is risky; unlimited usually means the studio is undercharging.

Scope creep: When the project's deliverables or requirements expand beyond the original brief. The largest cause of timeline and budget overruns in AI production. Best mitigated by tight initial briefs and explicit aspect ratio variants in the original quote.

Master file delivery: Final delivery of the master deliverable, typically as a high-quality file plus relevant aspect ratio variants. Standard project completion milestone.

Geography and economics

Cost arbitrage: The phenomenon of equivalent production craft being available at meaningfully lower cost in different geographies. Discussed in detail in Why Global Brands Are Producing AI Films in Southeast Asia.

Boutique tier: Studio engagement at $25,000-$80,000 per project, typically a 5-15 person director-led shop.

Mid-market tier: Studio engagement at $80,000-$250,000, typically a 30-80 person production agency.

Enterprise tier: Studio engagement at $250,000+, typically a top-shelf agency or holdco production arm.

Decisions you will need to make

Pure AI vs. hybrid pipeline: The structural choice between generating every frame or combining AI generation with live-action footage. Different production paths with different cost and quality profiles.

Single-model vs. multi-model pipeline: The choice between using one AI model throughout or casting per shot. Multi-model produces higher quality but requires more sophisticated production capability.

Master + variant strategy: The decision about how many aspect ratio and length variants to commission upfront versus add later. Locking variants up front is much cheaper than adding them post hoc.

Studio vs. self-serve tooling: The decision about whether to engage a production studio or attempt internal production with self-serve tools. Often misanalyzed — see The Hidden Costs of Cheap AI Video Generators for why.

---

If you are evaluating an upcoming AI production project and want help mapping the brief to the right production tier and pipeline, we run pre-production conversations before any quote — getting the structural decisions right early saves 20-30% on production cost. For specific tool comparisons, see Veo 3 vs Kling 2.0 vs Runway Gen-4. For the broader budget context, see AI brand film cost breakdown for 2026.

Or see our Services overview for the full lane of production work we cover.