Hentai and AI Art Glossary: Key Terms Explained
Adult anime art sits where two vocabularies meet: Japanese manga and anime terms, and the technical jargon of AI image generation. This glossary covers both. Terms are grouped by topic and kept non-graphic. For background on the genre itself, see what is hentai.
Genre and culture terms
Hentai – In English, sexually explicit anime, manga or games. In Japanese the word means “transformation” or, colloquially, “pervert”, and is not the usual industry term.
Ecchi – From the letter H. Suggestive, playful or risqué content that stops short of being explicit. Also used in Japanese as a mild word for “sexy” or “naughty”.
Seijin-muke – “For adults”. A common Japanese label for 18+ manga, anime and games.
Ero manga – Japanese term for adult manga, used by publishers and stores.
Eroge – Adult computer games, often in visual novel format.
Visual novel – An interactive story game told mainly through text, character portraits and background art, with choices that branch the story.
OVA – Original video animation. Anime released directly to home video rather than broadcast. In the 1980s and 1990s this format allowed adult anime to exist outside TV rules.
Doujinshi – Self-published works, often fan comics, sold at events such as Comiket. Many are original works; many use existing characters. Buy them from legitimate sellers, not scan sites.
Mangaka – A manga artist and author.
Shunga – Erotic Japanese woodblock prints and paintings, especially from the Edo period (1603–1868). An important ancestor of modern adult manga art.
Tentacle trope – A fantasy motif in adult manga and anime, often traced back to Hokusai’s 1814 print The Dream of the Fisherman’s Wife.
Futanari – A fantasy genre label for adult characters drawn with both male and female sexual characteristics.
Ahegao – An exaggerated, comic-style facial expression convention in adult manga and anime: rolled-back eyes, flushed cheeks and an open mouth, meant to signal being overwhelmed. Widely parodied in meme culture.
Kemonomimi – Characters with animal features such as ears and tails on an otherwise human body (fox, wolf, cat). In adult art, these must be clearly adult characters.
Monster girl – A fantasy genre of hybrid female creatures: lamias (snake), harpies (bird), centaurs and so on. See anime genres overview.
Censorship mosaic – Pixelation, blur or bars over explicit areas, required for commercial adult content in Japan under obscenity law.
Art and drawing terms
Cel shading – Colouring with hard-edged shadow shapes instead of smooth gradients, named after the transparent celluloid sheets (“cels”) animation was once hand-painted on. Also a 3D rendering technique that imitates that look.
Screentone – Printed dot or line patterns used in black-and-white manga to create grey tones and textures. Now usually applied digitally.
Line art – The ink outlines of an illustration, before colour.
Line weight – The thickness of lines. Varied line weight gives depth and emphasis; uniform thin lines look “digital”.
Key visual – A polished promotional illustration for an anime or game, usually with dramatic lighting and composition.
Rim light – A bright edge of light outlining a character from behind, separating them from the background. See colour and lighting.
Character sheet – A reference page showing a character from several angles, with colours, costume details and expressions, used to keep them consistent.
AI image-generation terms
Diffusion model – The type of AI behind most image generators. It learns to remove noise from images and generates new ones by denoising random static under the guidance of a prompt.
Checkpoint (model) – The main trained AI model file. It determines the overall style the generator can produce, for example anime, semi-realistic or photographic.
LoRA – Low-Rank Adaptation. A small add-on file that adjusts a checkpoint towards a specific style, concept or original character, applied with a weight (usually 0.5–1.0).
Prompt – The text description you give the generator.
Negative prompt – A list of things the model should avoid, such as blur, extra fingers or text.
Seed – The number that sets the starting random noise. Same seed and same settings produce the same image, which makes it essential for controlled iteration.
CFG scale – Classifier-free guidance. Controls how strongly the image follows the prompt. Too low gives vague images; too high gives harsh, oversaturated ones.
Sampler – The algorithm that performs the denoising steps (for example Euler a, DPM++ 2M). Different samplers give slightly different looks and need different step counts.
Steps – How many denoising passes the sampler runs. More is not always better; 20–30 is typical.
img2img – Image-to-image. Generating a new image from an existing one plus a prompt, with a “denoising strength” that controls how much changes.
Inpainting – Regenerating only a masked area of an image, for example to fix a hand or change an expression. See editing and upscaling.
Outpainting – Extending an image beyond its original borders.
Upscaler – A model that enlarges an image while adding plausible detail, rather than simply stretching the pixels.
ControlNet – An add-on that guides generation using a pose skeleton, depth map, line drawing or other structure, giving precise control over composition.
VAE – Variational autoencoder. The part of the pipeline that converts between the model’s compressed working space and the final pixels. A wrong or missing VAE often causes washed-out colours.
Hires fix – A two-stage process that generates at native resolution, then upscales and refines, avoiding the duplicated limbs you get from generating large images directly.