A training video is great for showing a task. But people searching your help center usually want to skim: find step 4, copy a setting, move on. The best help-center pages have both, the written steps with a picture of each, and the video above them for anyone who'd rather watch.
Writing that article by hand means pausing the video, taking screenshots, blurring them again and retyping every step. This guide shows how to get most of it from the video itself, and what to check before you publish.
The example is a sales team's "Creating a quote" lesson, recorded in a made-up CRM.
1. What a good help-center article has
Whatever tool you publish in, the articles people find useful share a shape:
- A title that's the task, the way people search for it: "How to create a quote", not "Quotes overview".
- One or two sentences up top saying what the reader will be able to do.
- Before you start: the role or permissions needed, and anything that must already exist.
- Numbered steps, one action each, with a picture of the screen where it helps.
- The result: what the reader should see when they're done.
- Common questions: the two or three things people actually ask about this task.
- An owner and a review date, so someone notices when the software changes and the article goes stale.
A well-made training video already holds most of this. It's in the right order, it has a step for each action, and every screen is in it.
2. Start with a video that has clear steps
The written article is only as good as the steps behind it. In GlideClip:
- Record the task, one task per video, narrating as you go.
- In Make it safe, click Prepare video, then check every item in Review blurs.
- In Studio > Callouts, add numbered step cards ("STEP 1", "STEP 2"...). Suggest callouts can draft them from your narration and your script.
Each numbered step card becomes one step of the article, so give them short, action-first titles: "Start a new quote", "Add the product", "Apply the contract discount", "Submit for approval". The guide How to add callouts and highlights to a screen recording covers this part in detail.
3. Write each step up in AI data
Click AI data at the top of the rail, then Segments. There's one segment per step card.
For each one:
- Check the Task name and the times.
- Write the Answer-style summary: one or two sentences that answer "how do I..." on their own. Draft from the narration gives you a starting point from what you said.
- Fill in Actions, one per line. These become the numbered actions under the step.
- Under Questions this step answers, click + Add a question for what people really ask, such as "Can I change the discount after submitting?"
- Pick the step's picture: move the playhead to the right moment and click Use the playhead, or leave it on Automatic.
With Pro, Draft with AI fills in empty summaries, actions and questions for every step at once. It works from text only, with everything the scan found replaced by [REDACTED]; your typed text is never replaced, and anything AI wrote is marked so you read it before exporting.
4. Add the context a reader needs
A video can jump straight into the task. An article can't. Open AI data > Video details and fill in:
- System and Version, such as the CRM and its release, so readers know whether the article applies to them.
- Who it's for, such as "sales reps".
- Before you start, such as "a customer account with a contract price".
- Result, such as "the quote shows Pending approval".
- Owner and Review by, so the article has someone responsible and a date to check it again.
- Video link: the address where the video is published, for example on YouTube. Each step then gets a "Watch this step" link that opens the video at that moment.
5. Export the bundle
Open AI data > Export, fix anything listed under Before you can export, and click Export bundle (.zip) (a Pro feature). It's saved on your computer. Inside:
- guide.md: the article. The title, who it's for and what you need first, then one heading per step with its summary, numbered actions, questions, the "Watch this step" link and its picture, and the result at the end.
- screens/: the picture for each step, blurred like the safe copy. Tick Draw highlight boxes on the step pictures to show which button each step is about.
- segments.jsonl, captions.vtt and manifest.json: the same steps as structured data, the narration as captions, and a record of where it all came from.
Unzip it somewhere and keep guide.md and the screens folder together: the article points to the pictures inside that folder.
6. Put it in your help center or wiki
guide.md is written in Markdown, the plain-text format most documentation tools understand.
- If your tool imports Markdown files, import guide.md (some tools take the whole unzipped folder, so the pictures come along). Notion, many wikis and Git-based docs work this way.
- If your editor converts Markdown when you paste it, paste the contents of guide.md, then upload the pictures from screens/ into each step.
- If neither, open guide.md in any Markdown preview (code editors such as VS Code have one), copy the formatted result into your editor, and add the pictures.
Then put the video at the top of the article, embedded or linked, so readers can choose to watch or read.
7. Read it like a reader before you publish
Treat the export as a strong first draft:
- Read every step against the software. Does each action match what's on screen? A summary drafted from your narration can pick up an aside you didn't mean as an instruction.
- Check every picture. Is it the right moment? Is everything sensitive covered?
- Check the masked text. Text is masked wherever the scan found something, but names are only masked when you listed them, and small or fast-moving text can be missed.
- Fill in what the video skipped. Exceptions ("if the product isn't listed, ask an admin") and what to do when something fails often live only in your head.
8. Keep the two in step
When the software changes, record the task again and export again. The video and the article come from the same steps, so they stay in step, and the Review by date reminds the owner to check.
GlideClip doesn't connect to your help center or publish anything for you: the files are yours to put wherever your team keeps its documentation.
Try it on a sample
GlideClip opens a sample recording of a made-up CRM, so you can add step cards and see the AI data steps take shape without recording anything first. Your recording is processed in your browser and never uploaded.