Broadcast
Editor
Saved faces
Voices
Asset packs
Asking the server…
Faces
Loading…
Ready: a screen is watching and can be heard, or is set to house audio (it moves, the house speakers play the sound). Open one with On screen…
Say: text through TTS
Push: an audio file
Text
0 / 5000
Voice
Voice name or speaker
Pause between sentences, ms
Seed (idle layer)
Temperature
i
How freely the voice is sampled, sentence by sentence. It changes the delivery, not the words you send.
Lower, 0.5 to 0.7:
steadier and more even from one sentence to the next, with fewer skipped or invented words. Too low (under about 0.4) can sound flat, or loop and stall.
About 0.9:
Miso's default. Natural variation.
Higher, 1.0 to 1.5:
more expressive but less predictable. Tone and pitch jump between sentences, and words are more often slurred or made up.
Service default:
sends no temperature, so the service (or the clone's pinned value) decides. Miso takes 0.1 to 1.5.
0.90
service default
Tidy speech: numbers as words, short sentences joined, even loudness, soft edges
Previous sentence as context (experimental)
Stream (only for a TTS faster than real time)
TTS service and extra fields
TTS URL
Extra fields for the TTS request (
-p key=value
): numbers and true/false go as themselves.
+ field
Copy
Speak
Stop
The text goes to the TTS service and nowhere else.
Audio file (WAV)
Seed (idle layer)
Copy
Play
Stop
Record to file
1080p (1920×1080)
720p (1280×720)
480p (854×480)
HLD portrait (1080×1920)
high
medium
low