FerroxLabs Voice Profile Bootstrap
Estimated time: 30-45 minutes
This workflow captures the author's writing voice into a reusable profile using the Ghostwrite Yourself method: 22 questions across six voice dimensions. It reads an initial hypothesis from reference material, runs the interview, synthesizes the profile, and validates it on a sample post before saving to voice-notes.md. Review gates sit after the interview, after synthesis, and after the sample.
Voice profiles built from the interview alone land around 60% accurate. With three or more existing posts or transcripts as reference, they reach 85% or better. Push for reference material early.
By the end you will have: a synthesized voice profile (rules, signature phrases, anti-patterns, tonal matrix), the interview responses, and a calibration sample, assembled into voice-notes.md.
When to Use
- User wants to capture their writing voice into a reusable profile
- User wants the profile other content workflows will apply
- User wants the voice validated on a real sample before saving
Workflow Steps
Step 1: Gather Reference Material (uses: content-voice-profile)
Ask the user what they publish (newsletters, threads, blog, podcasts) and roughly how much existing material exists to reference. Encourage them to share even a paragraph or, better, three or more existing posts or transcripts, since reference material raises accuracy from about 60% to 85% or more.
- Input: description of what the user publishes, existing material
- Output: summary of available reference material
- Key focus: secure reference material to lift accuracy
Step 2: Read an Initial Voice Hypothesis (uses: content-voice-profile)
From the reference material, produce an initial voice-style read as a starting hypothesis. State it with your reasoning, and make clear the 22-question interview will refine it.
- Input: reference material summary
- Output: initial voice-style assessment with reasoning
- Key focus: a starting hypothesis, not the final profile
Step 3: Run the 22-Question Interview (uses: content-voice-profile)
Run the Ghostwrite Yourself interview: 22 questions across the six voice dimensions, starting from the initial assessment. Capture the responses. Show them, then ask whether to revisit any question or proceed to synthesis. If the user wants to revisit, re-ask the named questions against their feedback, up to two iterations.
- Input: initial voice assessment
- Output: captured interview responses
- Key focus: complete, honest answers across all six dimensions
Step 4: Synthesize the Voice Profile (uses: content-voice-profile)
Synthesize the profile from the interview responses and the initial assessment: voice rules, signature phrases, anti-patterns to avoid, and a tonal calibration matrix. Show it, then ask whether to test on a sample post or refine the profile. If the user wants changes, revise against their feedback and show it again, up to three iterations.
- Input: interview responses, initial assessment
- Output: synthesized voice profile
- Key focus: rules, signature phrases, anti-patterns, tonal matrix
Step 5: Test on a Sample Post (uses: content-thought-leadership-post)
Write a sample thought-leadership post using the synthesized voice profile, as a calibration test. Show it and ask the user to read it: does it sound like them? If yes, ship the profile. If not, ask what is off and refine the profile against that failure, then re-test, up to three iterations.
- Input: synthesized voice profile
- Output: calibration sample post
- Key focus: a real sample that proves the voice reads true
Step 6: Save voice-notes.md (uses: content-voice-profile)
Assemble the deliverable: the voice profile, the interview responses for reference, and the calibration sample, and save it as voice-notes.md so other content workflows can apply it.
- Input: voice profile, interview responses, sample post
- Output: assembled voice-notes.md
- Key focus: a saved, reusable profile other workflows can load