NovelAI

Overview

NovelAI is a subscription web app for AI-assisted creative writing and anime-style image generation, made by Anlatan (a Delaware-registered company). It launched in 2021 as a text-generation tool for collaborative fiction and added image generation in October 2022. It is popular with hobbyist writers, roleplayers, and anime-art enthusiasts, in part because it imposes few content restrictions and leans on a privacy-first message (encrypted text storage, generated images reportedly not kept on its servers).

What it does

On the text side, the writer enters prose and the model continues the story on request. Output is steered by a memory field, an author's note, module presets, and a Lorebook: structured entries for characters, locations, and worldbuilding rules that the model automatically pulls in for consistency across a long story. The current high-end text model is Erato, based on Meta's Llama 3 70B and released in September 2024, alongside older in-house models. On the image side, NovelAI Diffusion (a custom Stable Diffusion derivative, latest V4.5) generates anime and illustration-style images from text prompts, with multi-character prompting, image-to-image, inpainting, vibe transfer, and post-generation tools (background removal, line art, colorize, emotion) on a purpose-built canvas. A customizable text-to-speech voice can narrate stories aloud. Throughout, the human prompts and curates: NovelAI is an assistant, not an autonomous agent.

Integrations & setup

NovelAI runs as a hosted web app at novelai.net with no setup beyond an account. It is not designed to embed into other products and does not advertise a marketplace of integrations; an unofficial REST API exists and community Python SDKs wrap it, but there is no broad first-party integration ecosystem. It runs on Anlatan's own hosted models (the Erato text model derived from Llama 3, and the NovelAI Diffusion image models).

Pricing

Freemium. A free Paper trial grants a fixed number of text, image, and text-to-speech generations. Paid tiers are Tablet ($10/month), Scroll ($15/month), and Opus ($25/month). Higher tiers add larger context sizes (Opus advertises up to 28,672 tokens), more monthly Anlas (the credit consumed by image generation: 1,000 on Tablet and Scroll, 10,000 on Opus), access to better text models, and, on Opus, free generation of normal-sized images under certain limits. Annual billing is discounted.

Best for / not for

Best for hobbyist fiction writers, interactive-fiction and roleplay users, and anime-art creators who want strong long-form story tooling (Lorebook, memory, modules) plus capable anime image generation with few content limits. Less suited to teams wanting marketing-copy workflows or brand controls (see Jasper or Copy.ai), to anyone needing a deep integration ecosystem or business deployment, or to users uncomfortable with image models trained on scraped art (NovelAI Diffusion was trained on the Danbooru dataset without artist consent).

Alternatives

Character.AI is a closer fit for conversational roleplay; Midjourney and Leonardo.Ai are general-purpose image generators (Midjourney is more painterly, NovelAI more anime-specialized); Jasper targets marketing copy rather than fiction.