When you paste text into Hold Your Voice, you get a score. That score is not arbitrary. It is the output of the Signal Engine, our analysis framework that breaks your writing into measurable voice dimensions and compares them against a profile built from your best work. Here is how it works.
Starting with your voice profile
Before the Signal Engine can analyze anything, it needs to know what your voice sounds like. During onboarding, you answer a few questions about how you write, what you value in communication, and what your brand personality looks like. You can also paste samples of your best writing.
From these inputs, the engine builds a voice profile. This profile is not a vague description like "friendly and professional." It is a structured map of specific dimensions: your typical formality level, your sentence rhythm patterns, your vocabulary preferences, your use of rhetorical devices, your directness, your warmth. Each dimension has a target range derived from your inputs and samples.
The dimensions we measure
The Signal Engine evaluates writing across multiple voice dimensions. Each captures a different aspect of how you sound.
- Formality spectrum. Where your writing falls between conversational and formal. This includes pronoun use, contraction frequency, and sentence structure complexity.
- Directness. How quickly you get to the point. Do you lead with context or lead with the conclusion? Do you hedge or state things plainly?
- Warmth and empathy. The emotional temperature of your writing. Human and relatable versus detached and clinical.
- Confidence. Your use of qualifiers, hedging language, and assertive versus tentative phrasing.
- Vocabulary signature. The specific words and phrases that characterize your writing. Your preferred terms, your avoided terms, your distinctive expressions.
- Rhythm and pacing. The patterns in your sentence lengths. Do you alternate between short punchy sentences and longer flowing ones? Do you use fragments for emphasis?
Sentence-level analysis
The Signal Engine does not just score your writing as a whole. It works at the sentence level. Every sentence in your text gets evaluated against your voice profile. Sentences that match your voice profile closely are marked as on-voice. Sentences that deviate significantly are flagged, with specific notes on which dimension drifted and in what direction.
This granularity matters. A paragraph might be mostly on-voice except for one sentence that slipped into overly formal language. A global score would mask that. Sentence-level scoring exposes it, so you can fix the specific problem without rewriting everything around it.
How scoring works
Your overall voice score is a composite. Each dimension contributes to the total based on how closely your writing matches your target profile on that dimension. The score is not about whether your writing is "good" or "bad" in some absolute sense. It is about whether it sounds like you.
A score of 90 means your writing is closely aligned with your voice profile across most dimensions. A score of 60 means there is significant drift on one or more dimensions. The engine tells you which dimensions are off and by how much, so you know exactly what to adjust.
Learning over time
Your voice is not static, and neither is your profile. The Signal Engine learns as you use it. When you accept a rewrite suggestion, that feeds back into your profile. When you consistently write in a way that deviates from your original profile, the engine can flag that as intentional evolution versus accidental drift.
This is important because brands evolve. The way you wrote a year ago might not match how you want to sound today. The Signal Engine accommodates that by treating your voice profile as a living document rather than a fixed standard.
Why we built it this way
Most writing tools give you binary feedback: this is good or this is bad. They check grammar, readability, and tone. But tone is the wrong frame. Tone is situational. Voice is persistent. A tool that says "your tone is professional" tells you nothing about whether the writing sounds like your brand specifically.
We built the Signal Engine to answer a more useful question: does this writing sound like you? Not like a generically good writer. Not like a professional. Like your brand, with your specific patterns and preferences. That is the signal we are measuring, and that is why we called it what we did.