Pasting a few writing samples into ChatGPT and asking it to "match my tone" works for about 10 to 15 messages before the model drifts back to its default voice. Custom instructions help, but they rely on how you describe your writing rather than how you actually write - and most people are terrible at describing their own style.
Noren, a macOS tool from a small startup, takes a different approach. You feed it 5 to 10 samples of your real writing per format (emails, tweets, blog posts), and it extracts concrete patterns: word choices, sentence structure, punctuation habits, how you handle concessions and analogies. The output is a portable Markdown file you can drop into the system prompt of ChatGPT, Claude, Gemini, or any local model.
The company's blog post walks through a five-step manual version of this process for anyone who wants to try it without the tool. The core argument is sound: a 50-line document of specific, observed patterns ("uses short sentences after making a big claim," "never starts with 'I think'") beats a vague custom instruction like "write in a casual, friendly tone" every time. The manual process takes hours of reading your own writing side-by-side and cataloging what you actually do versus what you think you do. Noren claims to compress that into minutes.
The tool is free to start and runs only on macOS. There is no pricing information yet for paid tiers. The voice profile it generates is a plain Markdown file, so there is no lock-in - you can edit it, move it between tools, or stop using Noren entirely and keep the profile.
For anyone writing AI-assisted content at volume - newsletters, social posts, client emails - a well-built voice profile is genuinely useful. The manual method in the blog post is worth trying even if you never touch the tool itself.