Voice
Each scene gets one clip: a generated line, or a recording you make in the browser. A scene lasts at least as long as its clip plus a short pause, 1.3 seconds unless you change Pause after each spoken line under Project. For shorts, try 0.3, and set each scene's At least low too, since a scene never ends before that.
Generated voice
Generate this line or Generate all lines (Voice tab), or from the command line:
pnpm kick formercial voice my-productGeneration uses edge-tts, which speaks with Microsoft's neural voices. It sends the spoken lines to Microsoft's speech service. Install it with pipx install edge-tts.
A clip is made again only when its spoken line, the voice or the rate changes. Generated clips are voice/<scene>.tts.mp3, and git ignores them; anyone can remake them.
Choosing a voice
Set the voice name under Project → Voice, or voice.tts in project.json. The default is en-US-AvaNeural. List the others with:
edge-tts --list-voicesNatural-sounding female voices include en-US-AvaNeural, en-US-EmmaNeural and en-GB-SoniaNeural. The rate is voice.rate in project.json, like -4% or +10%.
Writing lines for the ear
The spoken line is separate from the caption so you can write it the way it should sound:
| On screen | Spoken line |
|---|---|
@forinda/kickjs-db | KickJS D-B |
| SQLite | S-Q-Lite |
| TypeORM | Type O-R-M |
| The database, the migrations, and the types | The database... the migrations... and the types |
Listen to each line in the voice panel before rendering.
Recording your own voice
Under Record it yourself:
- Record: the browser asks for your microphone the first time.
- Stop, then listen.
- Retake until it's right, then Use this recording.
The recording is stored as voice/<scene>.rec.webm and used for that scene from then on. To go back, press Use generated voice. Recordings are kept in the project (git includes them), up to 50 MB each.
Record in a quiet room, close to the microphone, and leave half a second of silence at each end.