Voice-over and subtitles
Pick a voice, set speed and volume, subtitle it properly. Plus one common trap: choosing a non-Vietnamese voice makes the AI rewrite the whole script.
Choosing a voice
Nearly every tool has a Voice field. The list holds male and female voices across several tones, plus any voices you cloned yourself.
- Preview before choosing. A voice that suits news rarely suits storytelling, and vice versa.
- Some tools let you select several voices at once — useful in batch runs, giving each video a different voice so the output feels less uniform.
- Templates for the comic, explainer and narrated-video tools can carry a default voice, so you do not re-pick it each time.
The voice-language trap
A non-Vietnamese voice makes the AI rewrite the script
Choose a foreign-language voice and the AI reads that as a request for a video in that language, so it rewrites the entire script from the original article into the voice's language. The script you just edited by hand is discarded.
To keep your own wording, pick a Vietnamese voice. The system warns you in place, but it is easy to click past.
In tools where you write the narration yourself, we also compare the narration language with the voice language and warn on a mismatch — Vietnamese text read by an English voice comes out mispronounced throughout.
Speed and volume
| Setting | Options | Guidance |
|---|---|---|
| Voice speed | Slow · Normal · Fast | Short social video suits fast; explainer content that viewers must follow does better at normal or slow. |
| Voice volume | A multiplier from 50% to 200% | Leave it at 100% unless background audio is competing. Pushing it too high distorts. |
| Music volume | Set in the branding template | Keep it at 10–15%. Higher drowns the voice-over. |
Subtitles
Most social viewers watch with sound off, so subtitles are close to mandatory. Turning them off still produces the voice-over; it only removes the on-screen text.
| Setting | Options |
|---|---|
| Position | Safe bottom · Bottom · Centre · Top |
| Style | Karaoke · Word by word · Block · Highlight |
| Font | Six groups: modern, bold, dynamic, soft, playful, classic |
Use Safe bottom for vertical video
TikTok, Reels and Shorts all overlay their own interface across the bottom of the frame. Subtitles at Bottom end up hidden behind the platform's buttons and caption. Safe bottom sits clear of that zone.
Karaoke and word by word hold attention better in short video; block is easier to read for longer, text-heavy content.
Cloned voices
You can create your own voice from a sample recording and then use it like any other voice in the list.
- Upload a WAV or MP3 of roughly 10–30 seconds.
- The file must be clean, one speaker, no background music, no noise. The sample quality determines almost all of the clone quality.
- Give it a name and gender so it is findable later.
- Listen to the preview before using it on a real video.
The TVC opening voice is an exception
In the TVC tool, the opening recording does not pass through the synthetic voice — we use the real voice in that file to build a lip-synced speaker. Record it in the voice you want the audience to hear.