Operational voice
Your own voice
Built from the material you chose to provide. VoiceMark keeps the founding intake as the source of truth and re-derives the active profile when its cycle is due.
VoiceMark, by SameGround
VoiceMark is a bidirectional writing environment: a chosen voice shapes the draft, the same voice reads the result back as located craft observations, and you decide what enters the working text. Similarity is not quality; the judgment stays with you.
Voice home
Your operational voice and your switchable references are related, but not the same thing. The operational voice is re-derived from your immutable founding intake; a reference is the particular voice this piece is being read or written toward.
Operational voice
Built from the material you chose to provide. VoiceMark keeps the founding intake as the source of truth and re-derives the active profile when its cycle is due.
The explicit comparison voice for this piece. Change it without rewriting your operational profile.
A reading against the selected reference, never a mark or authenticity claim.
Start from scratch
No text or brief is required. Answer the intake, choose or edit a direction, approve the bible and map, then begin. You can take over, edit, save, pause and resume at any point.
Concept directions are generated planning material and count against the write allowance. No prose is drafted at this stage.
VoiceMark proposes; you decide. Select one direction, then edit any field before the bible and map are built.
Next, VoiceMark builds two compact rival architectures only. It does not expand either branch until you choose one.
These are compact planning candidates, not recommendations. Compare the material difference, assumptions and vulnerabilities, then choose one branch to expand.
The two-candidate comparison is decision-support overhead and does not debit the write allowance. Only your selected architecture is expanded into a full creation pack.
Everything below is editable. The approved version becomes the durable spine used for drafting and resume.
These are two different things. Your operational own voice is derived from your founding Intake and is versioned by VoiceMark. A named reference is a switchable author, target or character voice for a particular piece. Choose one action below; one action never silently performs the other.
Build/update my own voice saves a new operational Intake version only. Save as named reference adds a switchable library reference only. Naming a reference does not rewrite your operational own voice.
Already built your operational own voice? .
VoiceMark never releases private Draft Zero. The right-hand mirror reads the released draft or your supplied text; it does not certify authorship, quality or misconduct.
Working text
Keep the source visible while you read, generate, compare and decide. A voice-match is closeness to the named reference - not a quality score.
Continuous mirror
Choose the next motion. Results remain beside the source so the loop is visible rather than hidden behind a one-way generation box.
A plain reading of what the piece is doing, its shape and intent.
VoiceMark reads it back as located observations, with a voice-match closeness reading against your chosen voice, whether you wrote it or piped it from another model. Similarity, not quality. The judgment stays with you.
Draft from a brief, or clean an existing passage, your own or another model's output, written toward your own voice.
A single pass writes one draft at whatever length the brief calls for. To set a word target and have it written in beats, tick long-form.
VoiceMark never releases Draft Zero in an ordinary response. Choose this only when you need the private Draft Zero → examination → revision chain retained for later provenance-sensitive use.
Set the length here. It is written in beats, each one sized to stay within a single pass, and beats are added rather than any one of them running over.
Editing a character record saves your correction as authoritative and automatically re-reads the manuscript for dependent continuity observations.
A separate, opt-in reading on the AI-or-human question. It never changes the voice-match above. It reports a signal, not a verdict, and never a number. Authenticity signals misfire on real human writing, so treat anything here as one input, not proof.
Record what context is present before the reading. Unverified context cannot strengthen the result, and a formal-use intention is disclosed rather than authorised.
Create or continue work of any form and length in the voice you selected. VoiceMark holds the approved spine, future map, rolling state and reviewed text; you can write one beat, auto-draft, take over, edit, save, pause and resume.
Hold a roster of students, set the target voice they write toward and the exemplars to read against, then examine a student's writing and share the feedback. The student sees craft notes and a match, never the method.
Students, the target voice and exemplars are profiles from your voice library. Make them on the Change voice screen, then build a roster here.
Voice over time
See closeness readings in context, mark changes in the reference snapshot, reopen released work and export a deliberately limited process record. This is a reading for you to interpret - not a development grade.
Each bar is a reading against the exact named reference and basis used at that moment. Gold marks a changed reference snapshot; do not read across that boundary as one continuous score.
Metadata loads first. Open a run deliberately to retrieve its released draft or located observations.