Guide
English
Search the docs⌘K
5.2

Text to speech

Turning a piece of text into speech is the most common job in AI video voice-over. In DOTV you just select an audio node that is still empty and the input bar floats up underneath: write what should be read, pick a voice, hit send. This section covers what each item on the input bar is, how to adjust the whole block of settings — speed, pitch, volume — and what you can do with the take once it's out.

What's on the input bar

The audio composerType what you want spoken in the middle. The row below holds the model and the voice; on the right are the Settings gear and the character count.
The text to read goes in the middle, the model and voice sit on the row below, the settings and character count are on the right.
  • The text to read — just write it. You can drop in pauses and interjections (see 5.5). This bar synthesizes plain text, so @ references are off here — anything you referenced would only get read out loud.
  • Model — two of them: Minimax-speech-2.8-hd (the default, steadier) and Minimax-speech-2.8-turbo (faster).
  • Voice — the pill with the swap icon. Click it for the voice picker (see 5.3).
  • Settings — the gear on the right opens speed / pitch / volume and the rest (below).
  • Character count — "N/10000", right of the gear. One synthesis takes at most 10,000 characters.
  • ⚡ and send — speech is priced per character, so that number moves as you type.

Settings

Three blocks: Basics, Timbre and SFX.
Block What's in it Range
Basics Speed 0.5–2, default 1
Pitch −12 to 12, default 0
Volume 0.1–10, default 1
Timbre Pitch shift, Intensity, Timbre −100 to 100 each, default 0
SFX None / Spacious echo / Auditorium PA / Lo-fi telephone / Robotic None by default
  • Every row is a slider plus a number box, both bound to the same value — type into the box when you want an exact number.
  • Each block collapses on its own. Reset all in the top right puts every setting (SFX included) back to its default.
  • The panel also has a Language row: Auto by default, or pin it to Chinese / Cantonese / English / Japanese and so on. Picking an accented voice moves it for you — swap the voice without the language and the model will read Mandarin in a Cantonese voice.

Once it's done

  • The waveform on the card is the audio; click it to play in place (see 5.1).
  • To change it: select the node, the input bar is still there, edit the text or the settings and send again. That's a fresh read, and it replaces the old one.
  • If you only want to touch the audio you already have — cut a piece off, speed it up — you don't need a fresh read. Use trim and speed instead (see 5.6).
Try it in DOTVReading only goes so far: open DOTV and make an AI video with the feature this section covers.
Open DOTVDownload the desktop app
Previous
5.1 Audio nodes
Next
5.3 The voice library
On this page
What's on the input bar
Settings
Once it's done