scribbler.live

What It Is

Scribbler is an open-source AI and interactive Notebook platform that runs entirely in the browser. Using WebGPU, WebNN, ONNX Runtime Web, and the JavaScript ecosystem, it lets users run LLM chat, Stable Diffusion image generation, Kokoro TTS, machine learning, and data visualization locally in their browser. Its core positioning is not as a cloud Notebook, but as a local AI experimentation environment with “no server, no installation, no login, and no API key.”

Core Features and Privacy

Based on the collected information, Scribbler supports dynamic loading of libraries such as WebLLM, TensorFlow.js, ONNX Runtime, Transformers.js, Plotly, and D3, and provides 50+ ready-to-run Notebooks. Notebooks support JavaScript, HTML, CSS, and Markdown cells, can be saved as .jsnb files, and support URL sharing, GitHub loading/pushing, iframe embedding, and HTML/JS export. Privacy is its strongest selling point: prompts, images, and data stay on the local machine, with no logs, no training input, and no backend dependency.

Pricing and Ease of Use

The page clearly states Free forever, Open Source, and MIT licensed, with no cloud API or server costs required, making it highly cost-effective. In terms of usability, regular users can simply open app.scribbler.live and try it out; developers can also download or clone the GitHub project and self-host it via a static server, GitHub Pages, S3, or an isolated network. However, it is primarily based on JavaScript rather than Python, which may create a migration cost for traditional data science users.

Pros, Cons, and Limitations

Its advantages include extremely lightweight deployment, a clear privacy-focused architecture, offline/local operation, easy sharing, and coverage across multiple scenarios including LLMs, image generation, TTS, ML, and visualization. The limitations are that performance depends heavily on the local GPU, browser WebGPU/WebNN support, and model size. Browser sandboxing can restrict certain APIs, CORS resources, and cookies; Notebooks may also be lost if local browser storage is cleared. Output quality also depends on the locally loaded model and should not be treated as equivalent to cloud-based large model services.

Who It’s For and Access from China

Scribbler is suitable for developers, AI educators, students, researchers, privacy-sensitive teams, and users who want to quickly share browser-based AI demos. The collected text does not specify access conditions from China. Availability may be affected by the domain, GitHub, CDN, and model resources, so actual network connectivity should be tested. There are no explicit payment requirements, as the current information indicates that it is free and open source. Alternatives to consider include Google Colab, JupyterLite, Observable, and Hugging Face Spaces.