On this page
- 01The short answer
- 02Choose how to create the narration
- 03Map actions before writing a long script
- 04Add imported narration in Rushilo
- 05Generate narration with your own provider key
- 06Repair timing by finding the first mismatch
- 07Create captions from the finished voice track
- 08Check the exported recording
- 09Sources
- 10Cite this page
The short answer
Short answer
To add voiceover to an existing screen recording, keep the picture and narration on separate tracks, divide the recording into meaningful actions, and align a short narration segment with each action. You can import recorded audio or generate speech from a script. Check the timing against the visible result before adding captions and exporting.
Use the browser loudness checker on a short mono/stereo excerpt to estimate integrated LUFS and sample peak. Match your delivery requirements and remeasure the finished mix; sample peak is not true peak.
You do not need to record the whole demonstration again because the original has no narration. You do need an audio recording or a text-to-speech service, and time to match the words to the picture.
Choose how to create the narration
Record your own voice separately
- Useful when
- Your delivery, personal presence or pronunciation matters
- What you still need to check
- Room noise, sentence timing and retakes
Import an existing audio file
- Useful when
- Someone has already recorded or approved the narration
- What you still need to check
- Its version, usage permission and timing
Generate speech from text
- Useful when
- You want narration without recording a microphone
- What you still need to check
- Names, delivery, provider costs and every generated line
| Method | Useful when | What you still need to check |
|---|---|---|
| Record your own voice separately | Your delivery, personal presence or pronunciation matters | Room noise, sentence timing and retakes |
| Import an existing audio file | Someone has already recorded or approved the narration | Its version, usage permission and timing |
| Generate speech from text | You want narration without recording a microphone | Names, delivery, provider costs and every generated line |
Microsoft's voiceover instructions document recording, text-to-speech and importing audio as separate routes. The route you choose changes how you create the audio; it does not eliminate the need to review synchronization.
Keep a human recording when the speaker's identity or delivery is part of the message. Text-to-speech is an option for a task explanation, rather than a requirement for every screen recording.
This workflow is for editing an existing video on a computer. Rushilo edits on Mac and Windows; its native screen recorder is Mac only. A recording made with another capture tool can be the starting footage. Rushilo product and platform details
Map actions before writing a long script
Watch the recording and identify the moments a viewer needs to understand: the starting state, the action and the visible confirmation.
Write one narration segment per useful action. Avoid a single uninterrupted paragraph that must fit the whole recording. Short segments make a later UI or wording change easier to manage.
Here is a hypothetical 30-second account-settings demonstration:
0-5 seconds
- Visible event
- Account settings opens
- Example narration
- Open account settings and choose Notifications.
- Review point
- The destination is visible before moving on
5-12 seconds
- Visible event
- An email preference is changed
- Example narration
- Turn on the weekly summary.
- Review point
- The control is large enough to recognize
12-20 seconds
- Visible event
- Save is selected and confirmed
- Example narration
- Select Save. The confirmation shows that the preference is updated.
- Review point
- The confirmation remains visible
20-30 seconds
- Visible event
- The selected state is reviewed
- Example narration
- Your weekly summary is now enabled.
- Review point
- The end state agrees with the narration
| Picture interval | Visible event | Example narration | Review point |
|---|---|---|---|
| 0-5 seconds | Account settings opens | Open account settings and choose Notifications. | The destination is visible before moving on |
| 5-12 seconds | An email preference is changed | Turn on the weekly summary. | The control is large enough to recognize |
| 12-20 seconds | Save is selected and confirmed | Select Save. The confirmation shows that the preference is updated. | The confirmation remains visible |
| 20-30 seconds | The selected state is reviewed | Your weekly summary is now enabled. | The end state agrees with the narration |
The intervals are an illustration, not a fixed pace. If the action is already complete before a sentence begins, shorten the sentence, move it or revise the picture timing.
For a new demonstration, planning the script before capture can help. For footage you already have, use the actual recorded states as your constraint. Do not narrate a button or result that the recording never shows.
Add imported narration in Rushilo
Keep an untouched copy of the source and save a named project version before changing the edit.
- Import the screen recording and the narration file.
- Place the picture on the timeline and the narration on an audio track.
- Identify any original microphone audio. Lower or mute unwanted narration so the viewer does not hear two voices.
- Retain system sounds only where they help explain the action.
- Move and trim each narration segment against its matching picture.
- Listen across every boundary, including the first and last words.
Rushilo's release history documents timeline editing, separate audio work and named versions. It also documents voiceovers with your own provider key. These are available editing capabilities, not a claim that the application has automatically synchronized your particular recording. Rushilo release history
If background music masks speech, follow the separate music-ducking guide. Adding another voice track does not by itself balance the mix.
Generate narration with your own provider key
Rushilo supports voice generation through your own OpenAI, ElevenLabs or Google Gemini key. Configure the selected provider in Settings and use your connected agent to generate a short segment. Your provider receives the request and bills you directly. Rushilo setup
Start with a sentence containing the product name and any unusual terminology. Listen to that test before generating the whole script.
An example editing brief:
Save a version first. Use the configured voice provider to narrate the approved account-settings script in short sections. Place each section beside the matching action. Inspect the current clip timing before placing audio. Keep the original recording, avoid overlapping voices, and show me where the narration needs more picture time. Export a review file after I have checked the timing; do not publish it.
This is an illustrative instruction. It does not promise a particular voice, generation time, model availability or cost.
If a model offers pace controls, they can be useful for a small adjustment. A sentence that explains too many actions still needs rewriting. Listen for natural pronunciation rather than forcing every line into its original slot.
Local video storage does not make generated speech offline. Use the local versus cloud editing guide to separate the operations in your workflow.
Repair timing by finding the first mismatch
Compare the narration and screen action near the beginning, middle and end.
Every line arrives equally late
- Likely editing problem to investigate
- Narration starts too late
- Useful repair
- Move the narration track or segments
One sentence runs into the next action
- Likely editing problem to investigate
- That sentence exceeds its picture interval
- Useful repair
- Shorten it or allow more picture time
Early lines fit, later lines move further behind
- Likely editing problem to investigate
- Segment durations and picture timing diverge
- Useful repair
- Realign at action boundaries rather than shifting everything
Two voices are audible
- Likely editing problem to investigate
- Original narration remains under the replacement
- Useful repair
- Inspect and lower the unwanted audio
A changed sentence no longer fits
- Likely editing problem to investigate
- The replacement has a different duration
- Useful repair
- Recheck that section and the following boundary
| Symptom | Likely editing problem to investigate | Useful repair |
|---|---|---|
| Every line arrives equally late | Narration starts too late | Move the narration track or segments |
| One sentence runs into the next action | That sentence exceeds its picture interval | Shorten it or allow more picture time |
| Early lines fit, later lines move further behind | Segment durations and picture timing diverge | Realign at action boundaries rather than shifting everything |
| Two voices are audible | Original narration remains under the replacement | Inspect and lower the unwanted audio |
| A changed sentence no longer fits | The replacement has a different duration | Recheck that section and the following boundary |
These are diagnostic suggestions, not automatic error classifications.
Suppose a hypothetical line is 6.8 seconds long but its picture interval is 5 seconds. It exceeds the interval by 1.8 seconds. You can rewrite it, give that action more time, or divide the explanation into two steps. Generating the same words again does not ensure a shorter result.
Preserve the order of cause and result. The viewer should see the action and the confirmation described by the narration.
Create captions from the finished voice track
Settle the narration and major picture edits before doing the final caption pass.
If you generated a new voice track, captions from the old microphone recording may contain different words or timing. Use the final speech, then review product names, punctuation and synchronization.
W3C's caption guidance explains that captions include speech and meaningful audio information, and that automatic output needs review. A pasted script is not proof of correctly timed captions.
Use offline Whisper subtitles for a local transcription workflow, and subtitle timing diagnosis if the cues drift.
Check the exported recording
Play the exported video outside the editor from start to finish. Confirm that it contains the narration, that every explanation matches the visible action, and that the final result is readable.
Test the actual destination player if you deliver captions separately. Keep the source recording, final narration files and editable project so one future sentence can be replaced without rebuilding the whole demonstration.
If you are starting with a new recording rather than repairing an existing one, use the product-demo shot plan. To edit the footage with your own connected agent, follow Rushilo setup.
Sources
- Microsoft: add voiceovers to videosRecorded, text-to-speech and imported-audio routes. Checked October 14, 2026.
- Rushilo release historyTimeline/audio editing, named versions and own-key voice generation. Recording is Mac only.
- Rushilo setupUser-configured provider keys and direct provider billing.
- W3C captions guidanceReview automatic captions and include speech and meaningful sound.
Cite this page
Last verified: Oct 14, 2026
Patel, O. (2026, October 14). How to add voiceover to an existing screen recording. Rushilo. https://rushilo.com/blog/add-voiceover-to-screen-recording
About the author
Founder, Rushilo
Om Patel is the founder of Rushilo, a desktop video editor driven by the user's own AI, and of BigIdeasDB, a research platform that reads real complaints and reviews to help founders validate products. He writes Rushilo's research articles with the same method: primary sources, dated numbers and the words real users write.
