AI content terms in one place.

C2PA

Also known as: Content Credentials · Coalition for Content Provenance and Authenticity

Industry standard for cryptographically signing media with provenance metadata — who made it, with what tool, when.

C2PA is an open technical standard developed by Adobe, Microsoft, BBC, Intel and others. It attaches a tamper-evident manifest to any image, video or audio file describing how it was created and edited. For AI-generated content, C2PA records that AI was used, which model, and what prompts/inputs. Meta, TikTok and major platforms read C2PA to decide whether to show "Made with AI" badges. Brands working with creators increasingly require C2PA on deliverables to comply with the EU AI Act (2026) and similar regulations.

"Made with AI" label

Visible badge platforms attach to AI-generated content based on C2PA metadata or manual disclosure.

Meta (Instagram, Facebook), TikTok, YouTube and others now display a "Made with AI" or equivalent label on posts containing AI-generated imagery. The label can be triggered automatically (via C2PA provenance) or set manually by the creator. Failing to disclose AI content can result in reduced reach, post removal, or platform penalties — and is mandatory under the EU AI Act for many use cases starting 2026.

Character Bible

A reference set of facial and body landmarks used to keep a person consistent across many AI-generated frames.

In AI photography, a Character Bible is a curated dataset (15–30 reference photos) plus metadata describing immutable identity features — bonework, mole position, scar, hairline, eye spacing, body type. The pipeline uses this to produce shots where the same person appears recognizably across 50+ different scenes. Without a proper Character Bible, generic AI tools tend to drift — the "person" in shot 5 looks subtly different from shot 50.

Human-curated AI content

AI-generated content where a human selects the keepers and discards bad outputs before delivery.

Raw AI tools generate hundreds of variations per prompt — most unusable due to anatomical errors, identity drift, lighting issues or composition flaws. Human-curated AI content means a person reviews the full output set and ships only the keepers (typically 5–10% of generations). This is the difference between "200 AI photos delivered" and "10 hero shots delivered" — quality over quantity.

EU AI Act

European Union regulation classifying AI systems by risk and mandating disclosure for AI-generated content.

The EU AI Act entered force in 2024 with phased compliance through 2027. Article 50 requires that AI-generated or AI-manipulated content depicting real people, places or events be clearly labelled as such. For creators and brands operating in the EU, this means C2PA provenance or visible disclosure on AI travel content, AI lookbooks, AI ads. Non-compliance penalties reach up to €15M or 3% of global turnover.

Micro-influencer

Creator with 1K–100K followers, typically high engagement and niche-specific audience.

In contemporary creator-economy taxonomy, micro-influencers (1K–100K followers) drive higher engagement rates than macro/mega tiers and command 4-figure brand-deal pricing. They are the primary audience for affordable AI content services because traditional photo shoots ($1500–4000 per drop) eat their margin, while AI-generated lookbooks at $50–600 fit their economics.

AI Lookbook

Curated set of AI-generated fashion or product imagery used to launch a drop, season or collection.

In fashion, a lookbook is a visual catalog showing pieces styled together. AI lookbooks generate the imagery without a physical shoot — same model wears multiple outfits in multiple settings, all rendered. Cost compared to traditional fashion photography (typically $3000–8000 per drop with model + photographer + location): $400–600 for an AI lookbook with comparable visual quality.

Deepfake

AI-generated media depicting a real person without their consent, often used misleadingly.

Deepfakes are a subset of AI-generated content where someone's likeness is synthesized without authorization, usually to deceive viewers. Legitimate AI photography services like fakesme operate on the opposite principle: explicit consent, biometric release, and clear AI disclosure on every deliverable. The distinction matters legally (deepfakes may violate likeness rights, defamation laws) and platformwise (deepfakes are removed; consented AI content with C2PA is allowed).

LoRA (Low-Rank Adaptation)

Fine-tuning technique that teaches an AI image model to generate a specific person, style or subject consistently.

LoRA is a parameter-efficient fine-tuning method used to specialize large image models (Flux, Stable Diffusion) on a custom dataset — for example, 15 photos of one person. After training, the model can place that person in any scene with high consistency. Most professional AI photography pipelines use a per-client LoRA combined with ControlNet for pose and a Character Bible for landmarks.

ControlNet

Conditioning system that lets AI image models follow a specific pose, depth map or composition.

ControlNet adds spatial control to diffusion models. Instead of relying purely on text prompts, the model is also conditioned on a reference image (pose skeleton, edge map, depth map). This means a creator brief like "same pose as reference photo, but in Bali" produces consistent, intentional output rather than random poses each generation.