Claude’s New Plugin Clones Your Writing Voice
▼ Summary
– A new Claude Code plugin called Voiceprint creates a personalized “linguistic fingerprint” from a user’s actual writing samples to make AI-generated text sound like them.
– The plugin is based on Stylometry, a research field that uses statistical analysis of linguistic patterns like word frequency and sentence structure to identify authorship.
– It analyzes five writing samples across different emotional tones and prioritizes the user’s actual writing habits over their stated preferences.
– The tool includes a feature allowing users to ban specific AI-generated phrases they dislike, such as “let’s dive in,” and provides tailored guidance for different formats like blogs or emails.
– Voiceprint is available on GitHub, takes about twelve minutes to set up, and works with any tool compatible with the SKILL.md specification.
A new Claude Code plugin called Voiceprint offers a powerful way to personalize AI-generated text by creating a unique “linguistic fingerprint” from your own writing. This tool analyzes your personal style to ensure all output mirrors your authentic voice, moving beyond generic AI prose to deliver content that feels genuinely yours. The plugin leverages advanced stylometric analysis, examining real writing samples to capture the subtle patterns that make your writing distinct.
The technology is rooted in stylometry, a field dedicated to studying linguistic patterns through statistical analysis of writing style variations. Historically used for determining authorship, this science now powers the plugin to clone an individual’s written voice with remarkable accuracy. James Kemp, the Core Product Manager for WooCommerce and the creator of Voiceprint, detailed the process in a social media post.
He explained that the system begins by collecting five different writing samples. These samples should cover a range of emotional tones, such as casual, explanatory, excited, frustrated, and persuasive. The stylometric analysis then scrutinizes these texts, focusing on elements proven to differentiate individual style. This includes the frequency of function words, the rhythm of sentence lengths, and habitual punctuation use.
A critical step involves cross-referencing the analysis against any user-stated preferences. When a conflict arises between what a user thinks they write and what the samples actually show, the plugin prioritizes the empirical evidence from the writing. As Kemp notes, how you actually write consistently beats how you think you write. The process also includes a phase for rejecting common AI clichés, allowing users to ban overused phrases that feel inauthentic, like “let’s dive in” or “in today’s fast-paced world.”
The final output provides a practical toolkit: a compressed list of phrases to avoid, specific examples of the user’s voice drawn from their submitted writing, and tailored guidance for different formats like social media posts, blogs, and emails. Available on GitHub, Kemp states that crafting a complete voiceprint takes approximately twelve minutes. The plugin is designed to work seamlessly with any tool that supports the SKILL.md specification, broadening its potential applications for developers and content creators seeking a more personalized AI assistant.
(Source: Search Engine Journal)
