The short version
- Write your script and put [pause 1s] wherever you want an exact silence.
- Put [at 0:12] before a sentence that has to start at 0:12, for example at a slide change or a cut.
- Generate with a preset or a saved voice and download the WAV.
- Place the file at 0:00 in your video editor: every marked sentence starts on its mark.
Welcome to the quarterly update. [at 0:08] Slide two shows this year's numbers. [pause 2s] And that is all for today.
In this example the second sentence starts exactly 8 seconds into the audio, and there is a 2-second silence before the last one. Markers aren't spoken and don't cost credits. Try it in the studio: the + Pause and + Start at buttons under the text box insert them for you.
Exact pauses
| Write | Silence |
|---|---|
[pause] | 0.5 s |
[short pause] | 0.3 s |
[long pause] | 1 s |
[pause 2s], [pause 1.5 seconds], [pause 750ms] | as written, up to 60 s each |
<break time="1.5s"/> (SSML) | as written |
The pause is the gap between the last sound before it and the first sound after it, accurate to about a twentieth of a second. Pauses in your own language work too, for example [Pause 2 Sekunden] or （停顿2秒）.
Without markers, VoiceAct still pauses naturally: at commas and full stops as the voice reads, and at a blank line between paragraphs. Use markers when the length matters.
Start a sentence at a set time
[at 0:12] means: the text after this marker starts 12 seconds after the beginning of the audio. VoiceAct fills the time before it with silence. Other ways to write it: [at 1:02.5], [at 0:01:05], [at 75s], [@ 0:12]. Plain timestamps from a script or transcript work too: [02:34], [01:02:34], [00:09.25], and [0:15.25] or [0:15]. A reference like [3:16] stays text; write [at 3:16] for that time.
Leave enough room. If the speech before a marker runs past it, the next sentence starts right away instead, a little late. As a rule of thumb, English speech runs at about 13–15 characters per second, so 100 characters need roughly 7 seconds. Slower voices need more.
Workflow for video and slides
- Note the times of your cuts or slide changes in the edit, for example 0:00, 0:08, 0:21, 0:35.
- Write one block of narration per slide and put the matching
[at …]marker in front of each block. - Generate with a preset voice or one of My Voices, then download the WAV (best quality) or MP3, and the SRT subtitles: word-timed captions that line up with the audio (ready a few seconds after it).
- Drop the file at 0:00 on your timeline. If a block arrives late, shorten its text or move the next marker.
Limits
- Up to 30 markers per generation; pauses and start times together add up to 10 minutes of silence at most.
- A marker at the very start adds silence before the first word:
[at 0:30]first means the voice starts at 0:30 in the file. A marker at the very end is ignored. - Voice Design without a saved voice: VoiceAct locks the new voice from your first sentence or two, so markers in that opening part are skipped. For exact timing from the first word, save the voice first and generate with it.
- Every marker starts a new take, so the voice picks up its intonation fresh after it. Put markers between sentences rather than in the middle of one.
Troubleshooting
| Problem | Try this |
|---|---|
| A sentence starts later than its marker | The text before it is too long for the gap. Shorten it or move the marker later. |
| The marker was read aloud | Check the brackets: square brackets, for example [pause 2s] or [at 0:12]. |
| No pause at the start of the audio | Put [pause 2s] or [at 0:05] before the first word; other silence at the start and end is trimmed. |
| The pause sounds shorter than set | Long breaths count as sound. Listen for the gap between words; it matches the marker. |
New to choosing a voice? See how to design a voice from a text description.