Chat with Nook like any assistant. Ask for a video or a picture and it makes one on your GPU. No account. No credits. Nothing you say leaves your computer.
Windows 11 · about 6 MB · no admin rights · Mac preview
Open Nook and start talking: a small model downloads by itself in the background, so your first conversation is minutes away. When you ask for a video or a picture, Nook makes it on your GPU, and asks first if that needs a model downloaded.
Ask for a picture the way you would ask a friend. If the model it needs is not on your computer yet, Nook says which one and how big it is, and waits for your yes. Then it makes the picture on your GPU and shows it in the conversation. Every result is a file on your computer.
Ask for a picture and Nook writes a rich description of it, then makes it with Z-Image Turbo: sharp, realistic 1024 by 1024 images in eight steps. Not quite right? Say what to change and it makes another. Every picture is saved on your disk.
Ask for a video in the chat, or open Studio to direct clips yourself and queue several. Nook renders locally with Wan and its own realism add-on, without generation credits or added watermarks. A 2-second clip takes about a minute and a half on an RTX 4060; Train teaches the model a look from your own clips and photos.
Your conversations, prompts and everything Nook makes stay on your computer. The chat model and the video and picture models run here, so once they are downloaded Nook works offline. Nook connects for model downloads, updates and an optional daily usage count that excludes your messages and creations. You can turn that count off in Settings.
Ask for an explanation, a draft or an idea, then turn it into a picture or a moving shot in the same conversation. Your conversations and results are kept on your computer.
What will you
make next?
Nook’s chat model comes in by itself, about 1.1 GB. The video and picture models come in when you first ask for one, after Nook has said how big they are and whether they fit your GPU.
Start with Wan 2.1 1.3B for short clips. Explore larger Wan models when your hardware has room. Image-to-video support depends on the model.
Explore local AI video generation →Sharp 1024 by 1024 pictures from a chat message. A one-time 6.7 GB download that fits an 8 GB card; Nook asks before it starts.
Explore local AI image generation →A small, quick model that answers in seconds, downloaded in the background on the first start. It runs on your card, and on the processor while a video is rendering.
Find your hardware starting point →Try another prompt without spending another credit.
Your hardware supplies the compute.Your conversations and prompts stay on your machine.
The models run on your computer.Keep the images and clips you create without a Nook watermark.
Check each model’s licence for usage rights.Chat, videos and pictures in the same desktop app.
Download your models, then work offline.Nook runs the local engines behind the chat, the pictures and the videos, and picks the build that suits your card. Install the app and start talking. No admin rights or WSL needed on Windows.
Notes on local generation: which Wan models fit, how much GPU memory is available, and what happens when a model runs out of room.
See all notesShort answers to the things people ask first. Anything else: hello@usenook.ai.
Nook is an AI assistant that runs on your own computer. Chat with it to ask questions or draft text, and ask it for a video or a picture to have one made on your GPU. Studio lets you direct clips yourself, and Train teaches the video model a look from your own clips. Explore the video generator and image generator.
Download Nook for Windows 11 or the Apple silicon Mac preview from the download section. Install the app and start chatting: Nook downloads its small chat model by itself in the background. When you first ask for a video or a picture, it asks before downloading the model it needs. Model downloads are separate from the small app installer.
An NVIDIA GPU with 8 GB of VRAM is a practical starting point on Windows 11 for images and short clips with smaller models. Larger models, higher resolutions and longer clips need more memory. AMD and Intel support uses Vulkan where available. The Mac preview requires Apple silicon and macOS 13.3 or later; 16 GB of unified memory or more is recommended. Check each model’s requirements before downloading.
Nook chats with Qwen3 1.7B, makes pictures with Z-Image Turbo and videos with Wan 2.1 and Wan 2.2. Memory requirements vary by model; Nook tells you whether one fits your GPU before you download it.
Local generation runs on your computer; your prompts, source images and results are not uploaded for processing. Nook connects for model downloads and updates, and can send an optional daily count of tools used. That count excludes prompts and creations and can be switched off in Settings. See the privacy policy for all app and website data handling.
Yes, once the app and the model’s required files are downloaded. Local generation does not require an internet connection. Downloads and updates do.
Nook is free for personal use. Local image and video generation does not use generation credits, and Nook does not add a watermark to your output. Your own hardware supplies the compute. The app’s terms and each model’s licence apply.
Check both Nook’s terms of use and the licence for the specific model you use. Commercial rights vary by model and use case; generating locally does not override those terms. Review your output and any third-party rights before publishing.
Download Nook and start chatting on your own computer. Free for personal use on Windows 11, with a preview for Apple silicon Mac. No account or generation credits required.
On Windows, SmartScreen may pause the installer: choose More info, then Run anyway. On a Mac, drag Nook to Applications and open it; the preview isn’t notarized yet, so the first time choose Open Anyway in System Settings › Privacy & Security. Trouble installing? Write to hello@usenook.ai.
