Fully local & private
The models run on your device and your characters, scenes, and transcripts stay in a local database. No account, no telemetry.
A private character chat
that never leaves your machine.
privstory is a character chat that runs local models on your own computer. Create a character, open a scene, and chat while your GPU paints their portrait and the world around them, and build your own modules to shape how characters and scenes are made. No cloud. No account. Nothing leaves your device.
Lena Halvorsen
Writing…
Lena
The lamp only turns for ships that have already decided to come home. You. I can't read yet. So tell me: which way is your bow pointing?▍
Say something to Lena…
Everything runs on your own machine.
Local models, GPU-accelerated images, your own modules, and nothing sent to a server.
Local chat model
A local Llama-3 model role-plays your character with streaming replies, running entirely offline once downloaded.
Portraits & scenes
An SDXL image model paints each character's portrait and repaints the scene backdrop as the story moves.
Characters from a brief
Describe a character in a sentence and the model fills in their persona, appearance, and a portrait.
Scenes as sessions
Open a scene with an opening setting and pick up the conversation later. Each character can have many scenes.
Build your own modules
Create and import modules, a chat model, an image model, and the prompts that drive character and scene creation, then switch between them.
GPU-accelerated
Metal on Apple Silicon and CUDA on NVIDIA, with a CPU fallback. Only one model sits in memory at a time.
Works offline
After the one-time model download, everything works with no connection. The only later network call is an optional license check.
From a sentence to a scene.
Three steps, all on your own hardware.
Download a model
Import a configuration and download its weights once, with a progress bar. From then on it's local.
Create a character
Write a one-line brief. privstory generates the persona and appearance and paints a portrait.
Open a scene & chat
Set the opening situation and start talking. The backdrop repaints as the character acts.
Free to try. License to go unlimited.
The free version gives you one character and one scene. A one-time license removes the limits.
Free
No account needed
- One character
- One scene
- Full local chat & image generation
- Portraits & scene backdrops
License
Buy once, use forever
- Everything in Free
- Unlimited characters
- Unlimited scenes
- Move the license between devices
Add your license key under Settings → License in the app.
Common questions.
What is privstory?
privstory is a desktop app that runs an AI character chat entirely on your own machine. You create a character, open scenes with them, and chat while local models paint a portrait and evolving scene backdrops.
Is it really fully local?
Yes. The chat and image models run on your device, and your characters, scenes, and transcripts are stored in a local SQLite database. No account is required and no conversation data leaves your machine. The only network access is the one-time model download and, if you add a license key, a lightweight license check.
Which operating systems are supported?
privstory runs on macOS, Windows, and Linux as a native app. Apple Silicon uses the Metal GPU backend automatically. NVIDIA GPUs use CUDA when built for it, and there is a CPU fallback.
Is privstory free?
The free version lets you create one character and one scene, with the full local chat and image generation. A one-time license unlocks unlimited characters and scenes and can be moved between your devices. You enter the license key under Settings → License.
What are the hardware requirements?
A GPU is recommended. The default model set fits in roughly 12 GB (chat about 5 GB, portraits about 6.5 GB), and the two models are never resident at the same time, the chat model is unloaded before a portrait is painted. Apple Silicon uses Metal and NVIDIA uses CUDA; a CPU fallback works but image generation is slow.
Do I need an NVIDIA GPU on Windows or Linux?
The downloadable Windows and Linux builds run on the CPU by default. For CUDA acceleration on an NVIDIA GPU, build from source with the cuda feature (see the README on GitHub). On macOS, Metal acceleration is built in.
How do I move my license to another device?
The license is per-device but portable. In Settings → License on the old device, choose Remove license to free the seat, then enter the same key on the new device.
macOS says the app "cannot be opened", how do I open it?
privstory isn't notarized by Apple yet, so macOS blocks it on first launch. Double-click privstory, and in the warning click Done (do not move it to the Trash). Then open System Settings → Privacy & Security, scroll to the Security section, and click Open Anyway next to the message about privstory, then confirm with your login password. Afterwards it opens normally.
Windows shows "Windows protected your PC", how do I install it?
The installer isn't code-signed yet, so SmartScreen warns about it. When the dialog appears, click More info, then Run anyway. If that button doesn't show, right-click the downloaded .exe, choose Properties, tick Unblock in the Security section, click OK, then run it again.
Can I customize how characters and scenes are created?
Yes. privstory is driven by modules, shareable bundles of a chat model, an image model, and the prompts that generate characters and scenes. Build your own module or import one, download its weights, then choose it when you create a character. Each character stays bound to the module it was made with.