Glossary

This appendix's own annotation, not an honesty marker: field vocabulary covered at category level rather than named in a chapter.

Book conventions

The marker meaning the claim has no corroborating source you can check.

Book conventions Used in STYLE.md §6

A face-detail pass applied after upscaling; the book names the operation rather than this button.

Editing, conditioning, and post-processing Used in 5.5

Promoting someone else's product for a commission: the cheapest first dollar, the weakest position.

Distribution, monetization, and metrics Used in 9.1, 9.7

The transparency article: 50(2) marks output and binds the provider, 50(4) binds you.

Law, compliance, and provenance Used in 11.1, 11.3

Generated creative produced to a brand's brief and priced per deliverable, needing no audience of your own.

Distribution, monetization, and metrics Used in 0.3, 9.5

The human check immediately before the irreversible step: publication, or a render you pay per second for.

Pipeline, automation, and production Used in 7.1, 7.3

Average revenue per user per period; in the book's worked month, $4,230 across 60 subscribers, or $70.50.

Distribution, monetization, and metrics Used in 10.4, 12.5

The trained weights a persona's look comes from; its license governs any adapter trained on it.

Models, training, and identity adapters Used in 3.1, 4.5

One production run, sized from the survival rate rather than from ambition.

Pipeline, automation, and production Used in 5.4, 5.6

A paid deliverable for a company, quoted as units with usage rights and exclusivity written in.

Distribution, monetization, and metrics Used in 9.5, 11.3

Content Credentials: a signed manifest describing what created a file and what was done to it.

Law, compliance, and provenance Used in 11.5

The ten facts a persona cannot break, each specific enough to check in ten seconds.

Identity, canon, and characterization Used in 2.3, 5.6

Naming in text what an image contains; what you name is what separates from the trigger word.

Models, training, and identity adapters Used in 4.5

Classifier-free guidance: how hard the model is pushed toward the prompt; high values over-bake an identity LoRA.

Sampling and inference settings Used in 3.2

The six-part specification that makes a persona reproducible as a character across months.

Identity, canon, and characterization Used in 2.3, 2.7

One saved state of a model or adapter, identified by a hash and recorded as provenance.

Models, training, and identity adapters Used in 4.5, 11.5

The rate at which subscribers leave; one month less of average lifetime moved LTV from $352 to $585.

Distribution, monetization, and metrics Used in 10.4, 12.5

A CLIP-style text encoder that responds best to comma-separated fragments; used by older model families.

Models, training, and identity adapters Used in 3.2

The image-tower half of a CLIP-family encoder, conditioning on a reference image rather than on text.

Models, training, and identity adapters Used in 3.2, 4.3

The node-graph engine the book uses as a production engine, callable by an orchestrator.

Pipeline, automation, and production Used in 3.4, 7.2

Reviewing a planned post for who it could harm or mislead before publishing it.

Law, compliance, and provenance Used in 11.6

A conditioning path constraining generation with a structural map such as pose, depth, or edges.

Models, training, and identity adapters Used in 4.3

Monthly pipeline spend divided by posts published; below roughly $6 is healthy, above $15 means step down.

Distribution, monetization, and metrics Used in 0.4, 7.5

Generation spend divided by frames passing QA; the metric that makes tier changes comparable.

Distribution, monetization, and metrics Used in 5.6, 7.5

Access, rectification, erasure, portability, and objection under Articles 15–22.

Law, compliance, and provenance Used in 11.7

Designing rather than collecting the training set: score a generated pool for identity, then for coverage.

Models, training, and identity adapters Used in 4.4

Content resembling existing persons or events that would falsely appear authentic; the definition triggering 50(4).

Law, compliance, and provenance Used in 11.1

Under the AI Act, whoever uses an AI system under its own authority; you, every time you send a prompt.

Law, compliance, and provenance Used in 11.1

A face-detail operation run last at final resolution and low denoise; over-detailing reads as synthetic.

Editing, conditioning, and post-processing Used in 5.5, 3.2

The five ready fragments placed by one rule: unavoidable, not available.

Law, compliance, and provenance Used in 11.3

A data protection impact assessment, required for large-scale profiling such as a chat persona.

Law, compliance, and provenance Used in 11.7

Fine-tuning more of the model on a handful of images: higher fidelity, gigabyte outputs, more overfitting.

Models, training, and identity adapters Used in 3.2

Dubbing replaces the audio track over existing visuals; regeneration re-produces the clip, which the mismatch rule prefers.

Video, motion, and audio Used in 6.4, 6.7

The adapter reproducing whichever expression or pose dominated its dataset; a concentration failure, not a prompt failure.

Models, training, and identity adapters Used in 4.3, 4.4, 5.6

Frame-to-frame instability in brightness, texture, or detail; a generator limitation, not a grading problem.

Video, motion, and audio Used in 6.1, 6.7

The EU data protection regime, RODO in Polish; it governs chat logs, references, and consent records.

Law, compliance, and provenance Used in 11.4, 11.7

A bracketed claim label: [UNVERIFIED], [VENDOR CLAIM], [UNTESTED], [SNIPPET], or [AS OF 2026-10].

Book conventions Used in STYLE.md §6

The largest total line in the pipeline cost table, and the line operators omit.

Pipeline, automation, and production Used in 7.5, 0.4

Generating motion from a validated still; the still-first workflow is what makes keyframes portable.

Video, motion, and audio Used in 6.1, 6.2

The slow migration of a face away from the reference plate, invisible in one frame and obvious across a year.

Identity, canon, and characterization Used in 4.1, 4.6

The 1–5 per-frame rating; nothing scoring 2 or below survives, whatever the mean says.

Identity, canon, and characterization Used in 4.6, 5.6

Denoising a whole image from your input at a strength value, restyling a shot while keeping its framing.

Editing, conditioning, and post-processing Used in 3.2

Regenerating only a masked region; the surgical tool for a hand, an eye, or one bad detail.

Editing, conditioning, and post-processing Used in 3.2, 5.5

A face embedding plus a landmark path, so one reference holds identity as pose and scene change.

Models, training, and identity adapters Used in 3.2, 4.3

IPAdapter carrying an identity embedding extracted by a face recognition model, which sharpens retention.

Models, training, and identity adapters Used in 3.2, 4.3

Conditions generation on reference images, transferring style and composition as well as face; unreliable as identity alone.

Models, training, and identity adapters Used in 3.2, 4.3

The validated still a clip is generated from, carrying the identity the video model would otherwise sample.

Video, motion, and audio Used in 6.1, 6.2

The compressed representation the model denoises in, which is why img2img is approximate and inpainting is soft-edged.

Models, training, and identity adapters Used in 3.2

The same interest framed as a right in one's own image; in Poland, prawo do wizerunku.

Law, compliance, and provenance Used in 11.7

Matching synthesized speech to a face; priced per minute and the most retried line in the cost sheet.

Video, motion, and audio Used in 6.6

A small low-rank adapter, tens to a couple hundred megabytes, that nudges a base toward one face.

Models, training, and identity adapters Used in 3.2, 4.5, 4.7

Rank sets the adapter's capacity; alpha scales how hard it pushes. Capacity, not quality.

Models, training, and identity adapters Used in 4.5

Blended ARPU multiplied by average subscriber lifetime in months; $70.50 across five months gives $352.

Distribution, monetization, and metrics Used in 10.4

Provider-side marking as C2PA credentials or pixel watermarks; not your obligation and not reliable.

Law, compliance, and provenance Used in 11.1, 11.5

The one-page document answering a brand's capacity, compliance, and rate questions before they ask.

Distribution, monetization, and metrics Used in 9.5

A platform that sells on your behalf, remitting VAT and issuing the invoice in its own name.

Distribution, monetization, and metrics Used in 9.4, 11.8

The identity techniques ordered from cheapest and weakest to most expensive and most portable.

Models, training, and identity adapters Used in 4.2

Written permission from a human appearing in a reference, scoped by territory, term, and media.

Law, compliance, and provenance Used in 11.4

The parameter controlling how much movement a clip is asked for.

Video, motion, and audio Used in 6.2

The text describing movement alone, kept short when identity conditioning is heavy.

Video, motion, and audio Used in 6.2, 6.3

The written set of things the persona will not do, including political deepfakes.

Law, compliance, and provenance Used in 11.6, 12.4

A saved generation workflow expressed as connected nodes, runnable from an API or script.

Pipeline, automation, and production Used in 7.2

The automation layer that calls the engine, passes parameters, and files outputs; n8n here.

Pipeline, automation, and production Used in 7.2, 7.3

Extending an image beyond its original frame; the book composes wider at generation instead.

Editing, conditioning, and post-processing Used in 5.5

Overfitting memorizes the reference set including its lighting; underfitting leaves a face too weakly learned to hold.

Models, training, and identity adapters Used in 3.2, 4.5

A surface you control, such as an email list, so the audience survives a platform decision.

Distribution, monetization, and metrics Used in 10.1, 12.4

The numbered folder convention such as 00-bible/ and 40-output/ that keeps every asset addressable.

Pipeline, automation, and production Used in 3.6, 5.3

A unit of sound a lip-sync model maps to mouth shapes; difficult ones triple the retries.

Video, motion, and audio Used in 6.6

Pay-per-view: one-off paid content sold inside a subscription, priced at about $8 to $40 per item.

Distribution, monetization, and metrics Used in 9.4, 10.4

The maintained set of forty reusable prompt variants, refilled in the weekly review slot.

Pipeline, automation, and production Used in 5.2, 12.5

The recurring places and objects around the persona; what turns generation from invention into retrieval.

Identity, canon, and characterization Used in 2.3, 5.3

Generator, prompt, seed, references, timestamp, and checksum recorded per published asset.

Law, compliance, and provenance Used in 11.4, 11.5

Under the AI Act, whoever develops an AI system or places it on the market under its own name.

Law, compliance, and provenance Used in 11.1

A person's right to control commercial use of their likeness, separate from copyright.

Law, compliance, and provenance Used in 11.4, 11.7

The four destinations decided at QA rather than later: publish, bank, retouch, or discard.

Editing, conditioning, and post-processing Used in 5.6

Injects an identity embedding with contrastive alignment designed to leave the base aesthetic untouched; no training.

Models, training, and identity adapters Used in 3.2, 4.3

One minus the survival rate; the diagnostic that says a pipeline layer regressed rather than the model.

Pipeline, automation, and production Used in 5.6, 7.5

Frames passing QA divided by frames generated; the input to batch sizing.

Pipeline, automation, and production Used in 5.6

Replacing a face after generation: fast and cheap, weak across angles and motion.

Models, training, and identity adapters Used in 4.2, 4.7

Instruction-based edit models and adapters that hold the subject while the scene changes.

Editing, conditioning, and post-processing Used in 3.2, 5.1

The versioned plate of the persona's face and body that every generation is scored against.

Identity, canon, and characterization Used in 5.1, 4.6

Plain images of the class an adapter teaches, included to hold the base model's general behavior steady.

Models, training, and identity adapters Used in 4.4, 4.5

The twelve-defect diagnostic routing nine of twelve fixes back into the pipeline rather than into the frame.

Editing, conditioning, and post-processing Used in 5.6

The defined lifespan per data class: 12–24 months for logs, five years for invoices.

Law, compliance, and provenance Used in 11.4, 11.7

The earning routes ordered easiest to hardest, with operator hours attached to each.

Distribution, monetization, and metrics Used in 9.1

Payment as a percentage of what the other party earns rather than a flat fee.

Distribution, monetization, and metrics Used in 12.6

The identifier given to one pipeline run, making a re-run idempotent instead of duplicating outputs.

Pipeline, automation, and production Used in 7.2, 7.3

The algorithm walking from noise to image, such as Euler or a DPM++ variant.

Sampling and inference settings Used in 3.2

The frozen sampler, scheduler, steps, and CFG written into the persona's specification.

Sampling and inference settings Used in 3.2, 3.5

The noise schedule the sampler follows, such as normal or karras.

Sampling and inference settings Used in 3.2

The random starting point; the same seed with the same settings reproduces the same image.

Sampling and inference settings Used in 3.2, 3.5

Three frames at production settings, checked for identity, before every large batch.

Pipeline, automation, and production Used in 3.4, 7.2

The Polish limited company; the form the book's compliance material assumes.

Law, compliance, and provenance Used in 11.7, 12.4

The images a clip interpolates between; specifying both constrains motion and reduces drift.

Video, motion, and audio Used in 6.2, 6.4

How many denoising iterations run; full-quality bases converge near 20–30, distilled models near 4–8.

Sampling and inference settings Used in 3.2

A recurring access level priced for entry, mid, or high intent.

Distribution, monetization, and metrics Used in 9.4

Generating a clip from text with no reference frame; the weaker route for a persona.

Video, motion, and audio Used in 6.1

A long-context text encoder that handles full sentences and word relationships; write prose, not tag soup.

Models, training, and identity adapters Used in 3.2

A synced speaking clip; the format that moves a pipeline from Minimum to Working.

Video, motion, and audio Used in 6.6, 7.5

Whether appearance holds across the frames of one clip; the video form of drift.

Video, motion, and audio Used in 6.1, 6.3

The dated record of what the persona has already done or claimed, used to catch contradictions.

Identity, canon, and characterization Used in 2.4, 5.6

The token activating the adapter, and the token that absorbs whatever the dataset wrongly associates with the face.

Models, training, and identity adapters Used in 4.5

Synthesis of speech from text; the cheapest layer in the stack at solo volume.

Video, motion, and audio Used in 6.5

Generation from text alone with no reference; where a persona starts and the wrong tool for maintaining one.

Editing, conditioning, and post-processing Used in 3.1, 3.2

Raising resolution on pixels rather than latents, so it runs after composition, identity, and repair.

Editing, conditioning, and post-processing Used in 3.2, 5.5

The encoder-decoder pair between pixels and latent space; a mismatched one shows as color shift and washed shadows.

Models, training, and identity adapters Used in 3.2

One prompt-and-settings combination inside a batch, tracked so its survival can be measured.

Pipeline, automation, and production Used in 5.2, 5.4

The One Stop Shop: one registration and one quarterly return once the €10,000 threshold is crossed.

Law, compliance, and provenance Used in 11.8

Producing new speech from a sample of a real voice; anchoring to a fixed artifact is why it beats voice design.

Video, motion, and audio Used in 6.5

Describing a voice and having a model create one with no source recording: no consent problem, less consistency.

Video, motion, and audio Used in 6.5

The EU consumer's 14-day right to cancel, lost only when consent and confirmation are documented.

Law, compliance, and provenance Used in 11.8

The serialized node graph exported from the editor, which is what the orchestrator executes.

Pipeline, automation, and production Used in 7.2

Cloning from a short sample with no fine-tuning, at the cost of emotion-to-emotion stability.

Video, motion, and audio Used in 6.5