AI Porn Glossary: 30 Generation Terms Explained

AI content generation has grown its own vocabulary fast, and most of it gets used without explanation. This glossary defines the terms you'll encounter across Nyxa and our guides — each entry short enough to quote, complete enough to stand alone. Terms are grouped the way you'll meet them: writing prompts, judging image quality, choosing styles, making video, and understanding the safety rules. Throughout, one constant applies: everything generated here depicts fictional, invented 18+ characters, never real people.

Prompting Basics

Prompt

The written description a generator turns into an image or video — on Nyxa, the only input there is. A strong prompt names the subject (with an age anchor), appearance, wardrobe, setting, lighting and, for video, motion. The prompt guide covers structure in depth.

Negative prompt

A list of things the model should avoid, usually written after a --neg marker: --neg bad hands, blurry, watermark. It cleans up the classic failure modes — broken anatomy, artifacts, unwanted styles — and costs nothing to include.

Diffusion model

The class of AI model behind modern image and video generation. It starts from random noise and progressively refines it into a picture that matches your prompt. "Tuned diffusion pipeline" simply means a diffusion model specialized — in Nyxa's case for anatomy, lighting and motion.

Text-to-image / Text-to-video

Generation that starts from words alone: no source photo, no footage, no reference upload. Text-to-video adds the time dimension — Nyxa's video generator renders clips up to 10 seconds at 1080p straight from a prompt.

Seed

The random starting point of a generation. The same prompt with a different seed gives a different result; that's why regenerating the identical prompt produces variations rather than copies.

Style tag

A keyword that steers the overall look: photorealistic, anime key visual, octane render. Use one style vocabulary per prompt — mixing them makes the model split the difference badly.

Image Quality

Quality tags

Boosters appended to a prompt — 8k, sharp focus, volumetric lighting — that push resolution of detail, lighting sophistication and general polish. Two or three well-chosen tags beat a pile of ten.

Artifact

Any visible generation error: warped edges, melted textures, phantom objects, garbled text. Artifacts are the primary target of negative prompts.

"Bad hands"

The genre's most famous artifact — AI models historically struggle with fingers. It is standard practice to include bad hands, extra fingers in every negative prompt, which is why you'll see it in all of our library prompts.

Upscaling

Increasing an image's resolution after generation, typically with a specialized model that adds plausible detail rather than just stretching pixels.

Styles

Photorealistic

The style aiming for photograph-level realism — skin texture, believable lighting, camera language like 85mm portrait lens. The default register of the image generator.

Anime style / Key visual

The broad Japanese-animation aesthetic: clean lineart, expressive proportions, saturated color. A "key visual" is the polished promotional-art look — the single high-effort illustration that sells a show. See the anime generator.

Cel shading

Flat, banded shading imitating traditional animation cels, requested with cel-shaded, clean lineart. The opposite of the soft, continuous shading of realism.

Hentai style / Doujin

The explicit-illustration tradition of manga: bold ink, screentones, dramatic paneling. "Doujin" refers to self-published fan-scene works whose art style this register borrows. See the hentai generator.

2.5D

A middle ground between flat anime and full realism — anime proportions and color with rendered depth and lighting.

3D render / Subsurface scattering

The CG look of a game cinematic, requested with engine vocabulary: octane render, ray-traced reflections. Subsurface scattering is the effect of light penetrating skin slightly before bouncing out — the single biggest realism cue for rendered characters. See the 3D generator.

Waifu

A consistent original character you design once and return to across many generations — the same fictional woman in different scenes, kept stable by an appearance anchor. The waifu generator is built around this workflow.

Video

Temporal consistency

How well a character and scene hold steady from frame to frame. It is the central technical challenge of AI video — early models warped faces the moment anything moved. Good temporal consistency is what separates a usable clip from a glitch reel.

Frame drift / Morphing

The failure mode of temporal consistency: a face or outfit gradually changing identity across the clip. Countered with a fixed appearance anchor and negatives like flicker, morphing, inconsistent face.

Motion verb

The part of a video prompt that describes what actually moves and how fast: slowly turning toward camera, hair drifting. Without motion verbs you get a still image with a zoom.

Appearance anchor

A fixed block of character descriptors — hair, eyes, build, wardrobe — reused word-for-word across prompts so the same fictional character appears every time. The key to both waifu workflows and multi-clip videos.

Clip chaining

Making a longer sequence by generating several short clips with the same appearance anchor and cutting them together. The practical answer to per-clip length limits.

Interpolation

Generating in-between frames to smooth motion — the "interpolated smoothness" that makes 10 seconds of AI video look fluid rather than steppy.

Safety & Legality

Fictional character

A person who does not exist, invented from a written description. Every Nyxa output is a fictional 18+ character — the property that separates lawful AI adult art from real-person abuse. The legal explainer covers why this line matters.

Age anchor

An explicit adult age in the prompt (25yo), stating the character is an adult and helping the model render one consistently. Every prompt in our library carries one.

Face-swap

Replacing a face in existing media with someone else's — the technique behind most real-person deepfakes. Nyxa has no face-swap and no upload; the capability simply doesn't exist on the platform.

NCII / Deepfake

Non-consensual intimate imagery: sexual content depicting a real, identifiable person without consent, AI-generated or otherwise. Illegal or being criminalized across many jurisdictions, and blocked by design here — see the content policy.

Content filter / Blockers

The safety layers around generation: input-side filters that refuse prohibited prompts (real people, anything implying a minor) and output-side classifiers that check results. On Nyxa these are hard blocks, not suggestions.