create-voice

Build a voice profile from writing samples by running them through a 7-phase pipeline that extracts patterns and generates artifacts like profile.json and config.json. You need a minimum of 50 samples to start the calibration process.

Voice calibration is the process of analyzing writing samples to forge an authentic voice profile. You need it when creating a new voice, building a profile from scratch, or establishing a baseline for consistent authorship generation.

You need a minimum of 50 writing samples to create a reliable voice profile. The pipeline enforces this 50-sample minimum to ensure deterministic analysis and accurate pattern extraction for your voice artifacts.

Calibrating a voice from samples generates structured artifacts including profile.json, config.json, and SKILL.md. The pipeline converges your samples and metadata into these files during the Rule and Generate phases.

Yes, voice profiling uses a wabi-sabi principle to treat natural imperfections as features rather than errors. The pipeline extracts these stylistic patterns to ensure the generated voice profile maintains authentic characteristics.

To calibrate a new voice from scratch, progress through the 7-phase pipeline: Collect, Extract, Pattern, Rule, Generate, Validate, and Iterate. Each phase is gated to ensure deterministic analysis before proceeding.