Stop Slop StudioEdit Guard
A one-page guide from Stop Slop (Shane Mercer)
The short version: if you dictate your drafts, clean out the "um"s yourself. Don't hand the transcript to an AI to make it read smooth. In our tests, a light cleanup kept a dictated chapter reading as human-written. An AI polish of the same chapter got it flagged as AI.
| Test | What was done to the dictation | Result | Evidence |
|---|---|---|---|
| Raw dictation, ~1,100 words | Nothing | 100% Human | n=1 |
| Same dictation, script cleanup | Fillers removed, swears swapped, capitals fixed (this kit's script) | 100% Human | n=1 |
| First 300 words, AI polish | AI rewrote each sentence into smooth prose, same order | 100% AI | n=1 |
| Other dictations, light grammar fixes | Grammar, hyphens, punctuation | Human | n=2 texts |
These are small tests (n=1 or n=2), on one detector, on one writer's dictation. Treat them as leads, not promises. A light cleanup lowers the chance of a false flag. It doesn't guarantee any score, and nothing here makes AI-written text read as human.
Detectors read how sentences are worded and how they flow into each other. A script that only deletes filler leaves your wording and your rhythm alone. An AI polish swaps your wording for its own, even when it keeps your sentence order.
The free cleanup script (filler_clean.py, plain Python, no install) is a free download: filler_clean.py. Or do the same thing by hand with find-and-replace.
chapter3.txt.python3 filler_clean.py chapter3.txt -o chapter3_clean.txt --reportIt removes "like," "you know," "I mean," "bro," and "LOL", swaps swear words for mild ones (add --keep-swears to skip this), and fixes the capital letter at the start of each sentence. It never rewords, reorders or merges sentences. Edit the word lists at the top of the script to fit how you talk.
Save the raw transcript, the audio file if you have it, and the cleaned version. If your work is ever questioned, that history shows how the draft was made.
Stop Slop. Tests run by the Stop Slop studio on Pangram 4.1, Oct 2026. Small samples. Results can change when detectors update.