Neverending AI Storybook · published · build 25897282
Eight months of work. Nearly 300 changes. This is the biggest update the game has had yet. The goals were simple: make it easier to get started, make the stories better, and give you more ways to bring them to life, without giving up what makes this game different. Your stories can run locally on your own computer, with no required cloud services or subscriptions - but we have even more cloud options if that's the road you'd rather go down.
The entire game has been redesigned, with revamped illustrations alongside the prose, clearer choices, a light theme, and a dark theme. The new setup wizard checks your hardware and recommends models that fit your graphics card, so you spend less time wrestling with settings like a power-user and more time writing stories like you wanted.
The game now downloads models directly, with options for different hardware and quality levels. You can start small with lightweight models or try larger ones if your machine can handle them. Neverending AI Storybook uses llama.cpp, which supports dozens of model families, including Llama, Qwen, Gemma, Mistral, Phi, and DeepSeek. You can also bring your own compatible GGUF text model or safetensors image model. We've successfully tested model spanning across vram and regular ram up to 60+GB. The game will happily download models in the background while you play, and let you know the moment they're done downloading - with the option to apply them immediately on the next page load.
We've spent months testing long playthroughs and improving the story engine. Characters, items, and plot threads are tracked more carefully, so stories have a better chance of staying consistent as they grow. Repetition is caught and rewritten, outlines get more attention, and older details can return when they're relevant. Nine story presets have been rewritten, from Noir Mystery and Cosmic Horror to Cozy Bakery and Gothic Romance. Each has its own hook, stakes, and supporting characters with secrets of their own.
New narration/voice engine. Kokoro brings 28 voices, voice samples, and pronunciation for thousands of names.
More image options. Choose from Stable Diffusion 1.5, SDXL, and Flux, depending on your hardware.
New Configuration Wizard. Built in hardware detection recommends the best models for your hardware and helps free used VRAM.
Local video generation. Short illustrated story clips with music and ambience are now in testing. If you're interested in trying it early, feel free to check out our beta builds available through Steam.
Full controller support. Navigate the game from the title screen through story setup, gameplay, settings, and save management. We don't expect many users to care about this, but making a game that's accessible to EVERYONE that wants to play is very important to us.
Export your story as a PDF. By popular demand we've added support for exporting your story to PDF. Current support is very basic but will be refined soon.
Want to use a cloud provider instead? The game supports services including OpenAI, Claude, Gemini, OpenRouter, Groq, Together, Mistral, and DeepSeek, along with custom servers. Cloud services send your story and messages to another company. The game makes that clear, and local generation remains the default. We aren't really a fan of using cloud services, but I've heard your requests loud and clear. It's clearly something a lot of you want, and I'd rather see happy customers than force my views on others.
Thank you to everyone who's played, shared feedback, and helped make this game better over the years. I build this alongside my day job, and every bit of support means a lot. Here's to stories that go exactly the way we want and that never ever end.
See ya'll again before Christmas with video support for users with 16GB of Vram or more!