Calibrate the scan to your writing
This page covers the Your writing tab in Settings. It has two parts: your writing baseline, which tunes the scan to you, and the Repurpose lens, which gives Repurpose a style to aim for.
Your writing baseline
Without a baseline, PlainSpeak compares a rewrite with the draft you pasted. When that draft came from an AI model, the comparison protects the model's habits, not yours. A baseline fixes that: PlainSpeak measures pieces you wrote yourself and compares drafts with those.
Before you start
- Choose at least five pieces of the same form, totalling at least 4,000 words. Pieces you wrote before you used AI tools work best.
- Each comparison length needs five pieces that reach it. PlainSpeak compares at 150, 300, 500 and 800 words, so short samples can support shorter comparisons: five 200-word posts support the 150-word comparison only.
- Plain text (
.txt), Markdown (.md) or Word (.docx) files, one piece per file, up to 2 MB each and 20 MB in one selection. A Word file can be up to 10 MB, since images make Word files big. It is measured on its text, which is held to 2 MB like any other piece, and only its text counts towards the 20 MB. One selection can be up to 200 MB on disk in all. - All of one form: long-form, social or conversational. A selection that mixes forms is refused, with a count of each, and so is a piece too short to tell. A long-form baseline is only compared with long-form pieces.
Build it
- Open Settings, then Your writing.
- Under Your writing baseline, select Choose files and pick your files.
- Read the summary. It names how many pieces and words, the form, and which comparisons are on, at which lengths, with how many pieces each.
- Select Save.
If any file cannot be used, nothing from the selection is used, and the summary names each file and why. Any baseline you already had is kept. Save waits until the files have been read.
If some pieces sit apart from the rest, the summary names them. A co-writer, a tool or a different period can cause that. It is an advisory check, not evidence of who wrote them, and they are still included; remove those files and build again if you want them out.
Keep a copy, or move it to another computer
Download baseline numbers saves your baseline to a .json file: the numbers only, no text from your pieces. Load baseline file reads that file back, on this computer or another one, without your original pieces. A loaded baseline is used once you select Save, and a file PlainSpeak cannot read is refused whole, with the reason, leaving any baseline you had in place.
If saving fails because storage is full, the dialog stays open, and Download baseline numbers keeps the numbers safe until you can save.
What changes in the scan
PlainSpeak compares the opening of a draft, at the longest length your pieces support and the draft fills, with the openings of your own pieces at that same length. The line says which length it used, such as "comparing the first 800 words". The rest of a longer draft is scanned as usual but not compared with your writing.
- Two distance lines: how far the draft's opening sits from your writing, as a rank against your own pieces, such as "further from your writing than 4 of your 8 pieces". It describes where the draft sits among your pieces; it is not a verdict, and some of your own pieces sit in the furthest tenth too. That includes the pieces the baseline was built from: each one is ranked against the others, so your most unusual piece can land in the furthest tenth at a shorter length. At the length PlainSpeak judges it on, none of your own pieces should be far.
- Style measures judged against your range: hedging, intensifiers, passive voice, rules of three, abstract nouns, long sentences and serial commas are flagged when the opening goes clearly above your own usual range at that length. On a sample under 300 words that reading is shown as a note instead, because a short piece is often outside its writer's range for no reason at all; a note never counts toward Fix.
- Change from source, judged against you: after a rewrite, a measure that moved but whose opening stays within your usual range does not count. See Change from source.
A draft shorter than your shortest supported length, or of a different form from your baseline, is not compared, and the scan says why.
What changes in a rewrite
Your baseline calibrates comparisons and which general editing instructions apply. Your selected voice and Repurpose lens guide the rewrite. For a Polish or Rewrite of the same form, the scan's general edits, such as "cut hedging", are chosen against your range, and the rewrite is not told to keep the source's register where those edits move away from it. PlainSpeak does not imitate your writing and does not promise that a rewrite inside your range is a better rewrite.
Quotations, code and links in your pieces are left out of every count, as they are in a draft: a hedge you quote is not your hedge. A baseline built before 1.0.1 counted them, so after you read your writing again some ranges can be a little tighter, and a measure such as hedging is flagged slightly sooner. A baseline built before 2.1 needs reading again too, as abstract nouns are counted more fully.
To remove the baseline, select Clear baseline, then Save. If the summary says it was built with an older measurement version, the way PlainSpeak counts has changed since you built it: choose the same files again to use it. Until then it is kept but not used.
After an update that changes how PlainSpeak measures, your baseline asks to be read again. PlainSpeak keeps only the numbers, never your files, so it cannot do this by itself. A notice says so the first time you open the app after the update, and Settings and the scan say it until you act. Select Choose your writing again, choose the same files, then select Save, and your baseline is back. Until then the scan does not compare drafts with it. A baseline built by a newer version of PlainSpeak is kept the same way but not used here; choose the same files again to use it in this version.
Repurpose lens
A Repurpose has no source style to keep, so the lens gives it one to aim for.
- Under Repurpose lens, paste a post you rate.
- On the next line, type three dashes:
--- - Paste the next post. Three or four is plenty, each at least 15 words.
- Select Save.
The summary under the box shows the target PlainSpeak measured, such as the average sentence length. Only the measurements reach the model, as instructions; the posts themselves are not sent. The lens is used in Repurpose only. After a Repurpose, target difference shows how far the output sits from the lens, as a measurement, not a pass or fail.