Litplay uses the best model in the world for writing: Claude Sonnet 3.6. AI Dungeon may have many bells and whistles but Sonnet 3.6 is much smarter than their premium models.
The lore updates system develops story content behind the scenes to allow for more immersive and realistic storytelling. It functions like the DM's notes, preparing whatever awaits you. This is intended for murder mysteries or scenarios where there is important background information the player isn't supposed to be shown.
You can go anywhere with your stories. If you find yourself bored with the direction things are going, revise your previous actions and redirect the game in another direction. The continuity feature lets you try out as many different approaches as you want.
Litplay offers a free trial of all features. Simply log in and make an account to begin. Subscribe for continual access.
Do you want to be grasshopper? A grasshopper with a System? A sorcerer (with any spells)? A starship captain (with any ship)? A superhero (with any ability)? A wizard pretending to be a robot? A robot pretending to be a wizard? A mysterious being cursing artifacts, people or aliens?
Do you want to self-insert into almost any franchise? Sonnet won't know the most obscure ones but you can prompt it in.
Powerful tone control in our system prompt – write in the style of Homer's Odyssey, Tolkein, the Bible, Shakespearian iambic pentameter (or anything you can think of). Pick tones like supernatural horror or epic fantasy. Instruct the AI to take special care with stellar phenomena or carefully consider the consequences of you manipulating Newton's laws.
Humour too cringeworthy? Silliness too generic? Specify your desires exactly and enjoy fine control. You can change the system prompt ingame at any time ingame.
If the AI errs in some respect, you can enter worldwrit mode to directly edit the text, preventing further immersion-breaking mistakes.
Quickly load and edit prompts, take inspiration from the creativity of others.
While other AI adventure stories scrimp on token context, Litplay offers transparency and player control. When adventures get long, you can press the summarize button and start anew in a new context length + some of the last responses(you can of course edit the summarized start of the adventure if the AI summary is inadequate). Litplay never compresses the context length without your knowledge, though you will eventually want to summarize since long context lengths reduce quality.
If your prompting is too edgy for Claude, it'll respond with something like 'No Response'. Claude doesn't like sex or extreme, gorey violence. You'll need to reprompt. More commonly there are errors when the Sonnet API is overloaded - I can't do anything about this. If you see any game error messages, they probably are my fault. Ideally, please open browser dev tools (hotkey usually F12) and copy out what's inside, send me the info on what went wrong and what caused it.
Prompting is key. Strong violence is best done in the right context, in dark fantasy, military sci-fi or Warhammer-style settings. Specify the tone you want - bleak, grimdark, suitable for a 40K book...
In-context learning. The more you win in past iterations, the more likely you are to win in future turns. Sometimes Sonnet also shifts into an impressed mood and starts praising you for being clever in tactics. Alternately, if you specify for a difficult adventure then Sonnet will throw more and more opponents against you, it can get somewhat frustrating or feel cheap and arbitrary. You can moderate this with prompting, specifying for challenging but not unreasonable or contrived encounters.
Deepseek R1 is great fun and cheap but it doesn't perform so well for multi-turn games, it gets somewhat chaotic and incoherent. Plus you have to wait for it to think. I find Claude is preferable for long exchanges. I may end up adding Deepseek as another option though.
You get out what you put in. Brief, generic inputs get more generic outputs. Be creative, almost anything can be a prompt or story setting if you extend it out. You can mix two fictional settings together, add a twist, invent an alternate history, create your own setting and have Sonnet help you come up with story ideas... Litplay isn't limited to pure gameplay (though that's the intended purpose), you can ask other questions for brainstorming/idea-sparking.
I'm James Leeming, indie dev. I made Litplay after having a lot of fun RPing with Claude Sonnet 3.5 like it was a DM and I was the player. Sonnet has a huge knowledge base and it's fairly creative and smart. I really liked struggling with it, making tough prompts and then thinking creatively to beat the complications and problems it threw up against me. It's a fun mind-game! I also enjoyed wielding OP powers and managing crazy situations.
But there were some shortfalls. If Sonnet made a mistake in the webapp I couldn't fix it without regenning the whole response which felt wasteful. What if it's a good response but mixes up someone's name (this can happen if fake identities are being used)? As the context length rises it got dumber and more rigid, not to mention it chewed up my usage. It was bad at summarizing by just naively asking for a summary, I had to personally summarize stories and start a new session.
Sonnet also was making it up as it went along, there was a lack of structure and planning behind it, no consistency. I made Litplay (with features like lore updates, worldwrit and auto-summarization) to try and minimize the shortfalls with using Sonnet for RP/text adventures, while maximizing the upsides. Litplay is also about sharing human creativity. Much of the time I was thinking 'this is such a powerful AI but what can I use it for?'. Playing with other people's prompts can give you new ideas and make it more fun for everyone. The tagging/prompt saving system should also be better than saving your prompts in a Word document!