A voice for video: a narrator without recording or a mic

Voice-over is one of the most common speech-synthesis tasks: narrate a clip, a tutorial, an ad or a Reel without hiring a voice actor or recording yourself. A neural network reads the script in a voice that sounds professional — and you save the studio, the time and the nerves about your own voice on camera.

How video voice-over differs from a plain read

Technically it's the same voicing of text, but adjusted for video:

• Timing. The voice has to hit the edit — keep up with cuts and not lag behind. Sometimes the text is trimmed to fit the scene.
• Pace and pauses. In video, pauses do real work: they let the viewer take in a shot and land the accents.
• Intonation by genre. Ads — energetic, tutorials — calm and clear, storytelling — with mood.

How to make a voice-over track

1. Write the script in short phrases, meant to be spoken aloud rather than read with the eyes.
2. Pick a voice for the format (upbeat/calm, male/female).
3. Voice it — phrase by phrase if needed, to hit the edit more precisely.
4. Mix it with the video; add background music low in the mix if you like.

Type the voice-over script — get a finished narrator track. Your first generations are free after signing up.

Voicing in several languages

A big advantage of synthesis: one script is easy to voice in different languages with the same voice style. This is the basis of translating and dubbing video: transcribe the original, translate, re-voice. For reaching a foreign audience this is many times cheaper than hiring a narrator per language.

Common mistakes

• Phrases that are too long — the narrator "runs out of breath" and the viewer loses the thread. Cut them.
• Ignoring pauses — a solid stream with no breaths sounds unnatural and tiring.
• Voice mismatched to the topic — an upbeat ad tone in a calm tutorial is jarring.
• Forgetting the music level — the bed should be noticeably quieter than the voice.

What's next

You can voice things in another (synthesized) voice. Next — how to change a voice that already exists: shift the timbre, the gender, make it someone else.

In the Twelver chat you can get a narrator track for video right in the conversation — in the language and style you want. A few generations are free after signing up.