How to turn a written piece into a spoken video

Last updated 2026-08-16 · 5 min read

What is the short version?

If you have already written the thing, the video is a production problem rather than a creative one: someone has to read it well, something has to be on screen while they read, and the words need to appear as captions. Three separate jobs that only exist because you wrote something good.

1. Go to Vizard Agent and paste the text.
2. Say the language, the register, and who it is speaking to.
3. Say the length, the shape and whether you want captions.

What do you need before you start?

The finished text and a description of the delivery. Vizard Agent reads exactly what you paste, so the piece should be final — and the register matters more than usual here, because the same words delivered warmly and delivered urgently are two different videos.

• The text, final. Vizard Agent narrates the words as written, including the ones you meant to cut.
• The language. Any language works, and saying it explicitly matters for pronunciation.
• The register. Warm, direct, urgent, confiding. This decides the voice more than the words do.
• Who it is for. "Women in their thirties thinking about changing career" shapes the pacing and the pictures.
• The length and shape. Vertical and around a minute is the usual answer for this format.

What do you type into Vizard Agent?

Paste the whole piece rather than describing it. Everything else in the prompt is short: the language, the tone and the shape. Vizard Agent decides the visual treatment from the text itself, which is generally better than a visual brief written before the narration exists.

Prompt

Variants worth knowing:

• Your own voice. Upload a clean sample and Vizard Agent narrates in it.
• Typography instead of footage. "Animated text on plain backgrounds" is cheaper and often stronger for a written piece.
• Several backgrounds. Ask for the piece divided into sections with a different background for each.

What does Vizard Agent actually do?

Here is the order Vizard Agent worked in on a real piece narrated in Ukrainian. Two things stand out: it generated the independent pieces at the same time, and it listened to its own narration before doing anything with it.

1. Checks the working folder and the available templates.
2. Looks at the available tools and their prices before planning.
3. Loads the animated typography guidance.
4. Checks the narration and graphics parameters, then the background, music and subtitle generation options.
5. Generates the narration, the music and three atmospheric backgrounds in one parallel pass.
6. Checks the durations and previews the backgrounds.
7. Returns to the working folder and confirms what was produced.
8. Listens to the narration to check the pronunciation and the pace.

Step 8 is the one that matters in any language you do not personally speak. Generated narration goes wrong in small, specific ways — a stressed syllable in the wrong place — and the only way to catch it is to listen, which is what Vizard Agent did before building the video around it.

What does the result look like?

From the run this page is written from, probed on the delivered file: 1080x1920, H.264, 30fps, 69.1 seconds, AAC audio. Vertical and just over a minute, which is what the length of the text dictated rather than a target set in advance.

Ask for a shorter cut and Vizard Agent will tell you what has to go rather than speeding the narration up, because a rushed read is the fastest way to lose this format.

Expect tens of minutes rather than minutes. This category was not separately measured; comparable work runs a median of 28 to 38 minutes end to end, and across all projects the median cost by tier is Flash 47, Pro 55, Max 242, Ultra 263 credits.

When does this not work well?

The video is only ever as good as the writing underneath it, and generated narration has specific weaknesses that are easy to miss unless you are listening for them. Vizard Agent checks its own read, and a second pair of ears on anything that matters is still worth the two minutes.

• Pronunciation of names and places. Vizard Agent listens back and catches a lot, and you should still check any name that matters to you.
• Text that was written to be read, not heard. Long clauses and parenthetical asides work on a page and collapse when spoken. Read it aloud before pasting it.
• Backgrounds can undercut the words. Atmospheric footage under a serious piece can turn it into a perfume advert. Say plainly if you want restraint.
• A minute is about a hundred and fifty words. Longer pieces need to be cut or split, and cutting is a decision Vizard Agent should not make for you.
• Music makes anything sound truer. A confident read over a swelling bed is persuasive regardless of whether the argument holds. Vizard Agent will build exactly that, so it is worth reading the piece once without the music.
• The claims are yours. Advice about careers, money or health carries weight when it is spoken confidently over music.

How do you fix a result that came back wrong?

Name the line or the layer that is wrong. Vizard Agent keeps the narration, the generated backgrounds and the music as separate pieces, so re-reading a single sentence does not regenerate any of the pictures, and swapping a background leaves the read you already approved exactly as it was.

• "This word is mispronounced." Spell it phonetically and only that line is re-read.
• "Too fast in the middle." A pacing change on the narration.
• "Different background for the second section." One element regenerated.

How does Vizard Agent compare to doing it yourself?

The manual version is a text-to-speech tool in one tab, a stock footage site in another, a music library in a third, then an editor to assemble and caption it. Each step is easy. Doing all four for every piece you write is what stops people writing videos.

Common questions

Does it work in any language? Yes. Say which, and Vizard Agent generates the narration in that language rather than reading a translation with an accent.

Can it use my voice? Yes. Upload a clean sample of ten to ninety seconds and Vizard Agent narrates in it.

Will it change my words? No. It narrates what you paste, which is why the text should be final before you send it.

How long a piece can it handle? A minute is roughly a hundred and fifty words. Vizard Agent will narrate a longer piece, and past two or three minutes this format usually wants splitting into parts rather than stretching.

Can I get it as audio only? Yes. Ask Vizard Agent for the narration on its own if you want to build the video elsewhere or use it as a podcast segment.

Can it add captions in a second language? Yes. Say which, and whether both should appear on screen or only the translation.