Character.ai drops PipSqueak 2 and teases Lorebook

There is nothing more immersion-breaking in roleplay than an AI character suddenly forgetting its own lore or breaking personality fifty messages in.

Character.ai launched PipSqueak 2, a major model overhaul designed to replace the original PipSqueak engine. The new model is rolling out immediately to c.ai+ subscribers, while free-tier users will gain access in early May.

The announcement arrived with an honest acknowledgment from management. Character.ai recently hit users with tighter regional age gates, new feature caps, and more ads for free accounts. As a small team handling millions of monthly users without outside investors, the company explained that these unpopular moves were necessary to keep free access viable—and that this product drop is meant to make good on that promise.

Staying in character and showing its work

PipSqueak 2 focuses squarely on the top pain points flagged by long-time users.

First up is character consistency. PSQ2 significantly curtails personality drift, helping characters stay true to their defined personas even deep into marathon chats. Context retention also got a major lift, reducing the need to constantly re-explain established plot points.

Dialogue and prose saw noticeable polish as well. Characters produce sharper scene descriptions, sound less robotic, and fall into repetitive text loops far less often. Meanwhile, paying users get stability patches on the flagship DeepSqueak engine to cut down on gibberish and cutoff replies, with DeepSqueak 2 already in active training.

The on-platform Memory system received an overhaul too. The in-app page has been redesigned to track specific details like hairstyles, eye color, and subtle behavioral quirks. Chat threads now display real-time notifications whenever a detail gets logged to memory, and c.ai+ members will soon get a visual meter showing exactly how much memory capacity remains in a thread before compression kicks in.

Worldbuilders finally get Lorebook

The marquee reveal, however, was a preview of Lorebook—the platform's most requested feature to date.

Until now, creators building complex fictional universes had to cram massive world lore into character definitions or constantly feed exposition during roleplay. Lorebook changes that by letting creators build modular encyclopedias of factions, locations, and historical backstories attached to specific characters.

The model dynamically retrieves relevant lore entries on the fly depending on what the conversation touches. This canon persists across separate chats, making characters feel like they have actually read their own source material. The feature is currently in creator testing and will roll out first to paying subscribers before widening access.

Why it matters

For casual chatters, better memory means fewer annoying loops. But for roleplayers and interactive fiction writers, this update is a massive quality-of-life win. Being able to inspect what the AI remembers—and soon, knowing precisely when memory runs out via a visual gauge—takes the painful guesswork out of long-form storytelling.

Strategically, Character.ai is pushing deeper into specialized interactive fiction. While frontier models from OpenAI and Anthropic can roleplay, they lack dedicated tooling for persistent worldbuilding and character-specific lore management. Building Lorebook directly into the platform creates sticky retention that general-purpose chatbots cannot easily replicate.

We think Character.ai is playing a tricky balancing act quite smartly here. Squeezing free users with more ads and tighter limits naturally sparked backlash, especially for a platform that grew on freemium goodwill. Delivering substantial model upgrades and long-begged features like Lorebook proves the team is genuinely investing back into the product, giving dedicated fans a solid reason to forgive the ads—or finally pull the trigger on a paid subscription.