Video Voiceover
A voice over for YouTube videos turns a script you have already written into narration you place against your own edit. Paste the script, choose a voice, direct how it reads, and export a clean audio track.
Then do the thing that actually grows a channel: take the finished narration into another language.
The job
You have an edit. Maybe a tutorial, a product walkthrough, a faceless channel video. The visuals are locked or close to it, and now it needs a voice. Recording it yourself means a quiet room, a decent microphone, and doing the whole take again when you fluff line six.
This is the video sitting unfinished in two places most often.
A tutorial or product walkthrough where the screen recording is done and the steps just need a clear voice over them.
A faceless channel that runs on narration, where a consistent voice is the format, not the presenter.
For a lot of people one of those three things is missing, which is why the video sits unfinished.
Finished example
A 60-second tutorial segment, one voice, with the narration track shown against the timeline. The script was written to picture and generated in one pass, with pauses placed where the screen recording needs a beat.
The script was written one block per beat, so the narration lands on the cuts. In the second version one line of beat 3 was rewritten and only that beat was re-generated. Nothing else in the track moved.
Other languages
Once the narration works, dub it into 30+ languages rather than rewriting and re-recording per language. The catalogue covers 84 base languages, and 160 including dialects.
Dubbing runs on one clip at a time, up to 50MB or 10 minutes. A longer video is dubbed in sections. That is a real limit and it is better stated here than discovered at upload.
This is the step most people skip, and for a channel it is usually the cheapest reach available.
One clip at a time. Over either figure, the video is dubbed in sections.
Quality varies by language. Audition the ones you need in the voice catalogue before you commit to a release.
How it works
Split the script into blocks that match your cuts. One block per beat of the edit. This is the habit that makes everything downstream easy.
Import the script from a document or a URL if you already have it, in .txt, .docx or .pdf.
One block per beat means one audio segment per beat, placed against the cuts rather than dropped as one long file.
Pick from 1,411 voices across three quality tiers, Emotional, Premium and Standard. For talking-head and tutorial work the Premium tier is usually enough. For anything performative, spend the credits on Emotional.
Set the delivery per block, either with one of 7 emotion presets or with a short written instruction where a specific line needs a specific read. The instruction is not tied to a matched sample of the voice you picked, so leaning on it hard can pull the line away from that voice. Use it where the reading matters more than the exact timbre.
Premium is usually enough for tutorial and talking-head work.
Place timed pauses where the screen needs room. Set pronunciation for product names once so every mention agrees.
Export as MP3, as a voice and music mix, or as SRT captions.
Paste Your ScriptOpen the config file ⏸ pause and paste your API key ⏸ pause from the SaaS dashboard.
Premium is usually enough for tutorial and talking-head work.
Open the config file ⏸ pause and paste your API key ⏸ pause from the SaaS dashboard.
Timing
It matches because you write to blocks. One block per beat means one audio segment per beat. You place them against the cuts rather than dropping one long file on the timeline and hoping.
Shorten the line, or open the gap in the edit. A synthetic read does not compress the way a human narrator speeds up to hit a mark, so the fix is the script or the cut.
Re-generate that block. You do not re-record the video's whole narration, which is the main practical difference from doing it yourself.
It does not read your timeline, detect your cuts, or place audio into your editor for you. You export audio and place it in whatever you cut in.
So why leave your editor at all? Because your editor does not dub. That is the whole argument on this page.
When to record yourself instead. When your voice is the channel. Personality-led video is built on the fact that it is you, and a synthetic voice trades that away for speed.
After the first pass
Clean the recorded audio with the voice isolator before you decide you need to re-record it.
Clone it from a 10 to 30 second sample with voice cloning, then generate from script.
Assign different voices per block. The multi-track editor keeps them separate.
Pricing
If your channel is monetised, start on a paid plan. Commercial rights come with paid plans, and the free plan is for personal and evaluation use only. Free is a way to hear the voices and judge the quality before you commit, not a way to publish.
The free plan is $0 and paid plans start at $19/mo. You spend credits, and for text to speech one credit is one word, so a ten-minute script of around 1,400 words costs about 1,400 credits.
FAQ
Yes. Paste a script and generate audio before signing up.
Commercial rights come with paid plans. Check the platform's own disclosure rules on synthetic voice separately.
Yes, cloned from a 10 to 30 second sample.
Yes, dubbing covers 30+ languages.
Yes, as SRT.
No, and nothing in this category does. You place one audio segment per beat.
MP3, a voice and music mix, or SRT captions.