Live evaluation workspace

Tune a voice.
Measure the result.

Change the script, voice, pacing, and output. Every generated take stays in this browser session.

01

Script

0 chars

Insert direction

02

Voice source

Default Saved Voice ID Instant clone

Use the house voice

Useful for hearing the model before choosing a specific voice.

03

Delivery

Calibrate

Speed 1.00× Volume 0 dB Latency mode Audio format

Model variation

Temperature 0.70 Diversity 0.70

04

Session takes

Newest first · cleared on refresh

No signals captured

Your generated takes and latency readings will appear here.