How to Create a Step-by-Step Guide with Scribe

November 15, 2026

Scribe is a browser extension that watches you perform a process once and generates a numbered how-to document with cropped screenshots, auto-written instructions and optional redaction. A 10-step guide that takes 25 minutes to write by hand comes out in about 4 minutes.

How does the recorder actually capture a process?

Install the extension, hit record, and work normally. Every click or typed field produces a step: a screenshot cropped to the relevant region, an arrow or box annotation, and a sentence describing what you did.

Stop recording when the task is complete. The draft appears in the Scribe dashboard within a few seconds.

Two behaviors matter:

  • It captures clicks, not reasoning. If you scroll to find a button, the scroll is not a step, but the resulting click is.
  • It captures form fields as steps but records the label, not your typed value, when the field is detected as sensitive or password-typed.

What do you need to decide before recording?

Three decisions that are painful to change after capture:

  1. Windows or full screen. Window capture is cleaner and crops tighter. Full screen catches dialogs that appear outside the window, which window capture misses entirely.
  2. Account state. Record as a demo account, not your own. Your real account may contain customer names or internal IDs that get redacted at the pixel level only when Scribe detects them — and it does not detect everything.
  3. Language and tone. Set it before recording; regenerating copy in a different tone costs a full re-process.

Why does the generated text sound wrong?

Because it describes the mechanics, not the intent. You get “Click Settings” where a reader needs “Open Settings to turn on two-factor authentication.”

A 10-step guide needs roughly 45 minutes of editing if it will be published to customers. That editing pass is four things:

  • Fix the first step. Add the prerequisite. “Open Settings” is useless if the reader is not logged in.
  • Rewrite for intent. Every step’s opening verb should say why, briefly. “Select Save to keep changes.”
  • Delete redundant steps. Scribe turns a dropdown selection into two or three steps. Merge them.
  • Add the failure case. Note where a reader can go wrong. Generated guides never mention errors.

How do annotations and redaction work?

Annotations: arrows, boxes, numbered badges and blur. Pick one style — the default red box on every step reads as a warning after the third one.

Redaction: Scribe can blur text fields it identifies as personal data, but detection is imperfect. Manually blur anything with a customer name, email or account number before sharing. Redaction is applied to the exported image; the underlying page was still captured at full fidelity.

What does publishing look like?

Three outputs: a shareable web page with a link, a PDF export, and an embed. The web page is the default and updates if you edit the Scribe afterward — which matters for version drift, since nobody goes back to fix a PDF.

If your product changes, editing the original Scribe and letting the link update beats maintaining a separate document. That is the strongest argument for the tool over manual screenshots.

Pricing runs on a free tier with a limited number of Scribes, and paid plans per seat in the low double digits monthly for unlimited captures and custom branding.

Where it breaks

Multi-app processes. Recording across a browser and a desktop app in one Scribe does not work — the extension only sees the browser. Split into two guides or use a screen recorder.

Anything that takes longer than a few minutes with waiting. Long polling actions produce screenshots of loading states.

And it cannot explain concepts. A Scribe documents a click path; it never explains why the workflow exists. Guides that need reasoning still need a human writing.

Who should use it

Support teams documenting common tickets. Onboarding teams writing internal SOPs. SaaS companies that need help-center articles on a weekly cadence.

Skip it for one-time explanations to a single colleague, and for anything involving desktop-only software.

Quick Summary

  • Scribe records clicks and generates numbered guides with cropped screenshots in about 4 minutes.
  • Window capture is cleaner; full-screen capture catches dialogs but needs cropping.
  • Budget ~45 minutes of editing per customer-facing guide — the auto text describes mechanics.
  • Web page links update when you edit; PDFs go stale immediately.
  • Redaction detection misses some personal data; blur manually before sharing.
  • Best fit: browser-based support and onboarding docs. Poor fit: desktop apps and multi-app workflows.
Tags: