The expensive take is rarely the one with a small stumble. It is the take that reaches the final click before anyone notices the wrong monitor, silent microphone, private notification, or unreadable interface. A polished demo begins before Record and continues after Stop.
Start with one observable outcome
Write the viewer's starting state, the action you will demonstrate, and the final state they should recognize. That three-part spine keeps the recording about a task rather than a tour of every control.
- Name the audience and what they already know.
- Use a small, believable sample project.
- Script the opening and closing sentence; outline the middle as actions.
- Decide which details belong in linked documentation instead of the video.
Prepare a desktop that can survive the take
Close unrelated windows, silence notifications, remove secrets and personal data, enlarge text if the final player will be small, and place the demo app in its starting state. Use a clean account or synthetic data when possible.
On Wayland, the desktop portal mediates which monitor or window is shared. The selection happens through the portal flow rather than through unrestricted application access to the desktop. See the primary source.
Verify picture, microphone, and system sound separately
Treat the screen, microphone, system audio, and camera as separate inputs. Confirm the intended screen or window, watch each enabled audio meter, make a short sample, and play it back. For a systematic audio diagnosis, use the linked no-audio checklist rather than changing several settings at once.
- Confirm the selected monitor or window.
- Check microphone and system-audio meters independently.
- Frame and light the camera only if it adds useful presence.
- Record and replay five seconds before the real take.
Use zoom and cursor movement as signals
Move the pointer to establish location, pause, then click. Zoom only when it resolves a real legibility or attention problem, preserve enough surrounding interface to keep orientation, and let the frame settle before the next action. These are editorial recommendations, not claims that every demo needs motion effects.
Edit for comprehension, not perfection
Remove setup time, dead searches, repeated attempts, and long waits while preserving cause and effect. Keep short pauses where the viewer needs to read. If a cut makes the interface jump to an unexplained state, add a brief bridge or rerecord the action.
Captions should include the spoken information and meaningful non-speech audio needed to understand prerecorded media. See the primary source.
Export and inspect the delivery copy
Choose dimensions and a codec that the destination accepts, then inspect the exported file rather than assuming the timeline preview represents it. FFmpeg describes output creation as a pipeline of selected streams, optional filtering, and encoding or stream copy. See the primary source.
- Watch the opening, every edit, and the final frame.
- Check caption timing and spelling.
- Listen on headphones and ordinary speakers.
- Confirm small text remains readable in the actual player.
How Magic Pages makes help-center tutorials
Magic Pages uses Screenix to explain tasks in its help center, including connecting a custom domain and configuring a newsletter sending domain. The goal is to help customers follow the steps in the product.
Read the Magic Pages help-center guide ↗
Sviatoslav says combining the recording, camera and styling saves the team time and gives them more freedom over the design. Screenix handles the screen presentation, while Reaper and OpenShot remain part of their workflow.
“The design side is where it really wins for us: background, frame, how the zoom behaves, cutting between shots for transitions. That makes the videos feel like real tutorial guides instead of a plain screen recording.”
Their workflow
- Screenix: screen and camera recording, backgrounds, framing, zoom and cuts.
- Reaper: audio work alongside the screen recording.
- OpenShot: finishing touches, including the intro animation.
Sviatoslav’s documented Linux setup is Fedora Workstation with GNOME on Wayland.
This example shows how a screen recording can become part of an existing help article. The video demonstrates the task, and the written guide gives customers a place to revisit the steps.
Pre-recording checklist
- The outcome fits in one sentence.
- The app starts in the correct state with synthetic or cleared data.
- Notifications and unrelated windows are closed.
- The intended visual source is selected.
- Every enabled audio meter moves and a sample plays back.
- Pointer, zoom, camera, and text remain legible at delivery size.
- The destination, access level, and requested response are known.
Sources and verification
Technical claims in this guide were checked against the upstream documentation below. Screenix-specific behavior was also verified against the desktop implementation current on August 31, 2026.
- 1. ScreenCast portal specification
XDG Desktop Portal · The session lifecycle, selectable source types, cursor modes, restore tokens, and PipeWire handoff.
- 2. Understanding Success Criterion 1.2.2: Captions (Prerecorded)
W3C Web Accessibility Initiative · Prerecorded synchronized media needs captions that convey dialogue and important audio information.
- 3. FFmpeg documentation
FFmpeg · FFmpeg's transcoding pipeline, stream selection, codecs, filters, and output handling.
- 4. Linux screen recording has no audio? Check these 7 things
Screenix · The current Screenix source-selection and preflight workflow.