ııı voice demoSTUDIO / 01Connecting to studio…

THE VOICE AUDITION STUDIO

Find the voice.
Tell your story.

A little room to experiment. Give your words a voice,
turn a few dials, and listen for the one.

01 The script

YOUR WORDS, OUT LOUD
0 words0 / 3,000
↳

Punctuation shapes the delivery. Short sentences create room to breathe. Start with a paragraph and listen closely.

ON THE MIC

Kokoro · 82M

The model behind Ghost Story. Local synthesis, preset voices, and a little space to make them your own.

NO API FEESRUNS ON YOUR SERVER

02 The voice

QUICK START
UnhurriedUpbeat
LowerHigher
Blend a second voice ＋

An experimental mix of voice styles. Blends stay within the same accent.

Pitch and treatments are audio effects. These models don’t follow emotion prompts or clone a voice.

Ready for a sound check?

24 kHz WAV · up to 3,000 characters per take

LISTEN. TWEAK. REPEAT.

Your takes 00

Your list stays in this browser. Download favorites: audio expires after 24 hours or 200 newer takes.

♫

Something good starts with a first take.

Generate your script, then compare voices side by side.

BEYOND THE FIRST TAKE

Other voices worth exploring.

Open-source options for the next experiment.
Only Kokoro and Piper are installed in this studio.

EXPRESSION + CLONING

Chatterbox

MIT licensed. Original Chatterbox offers exaggeration controls; Turbo adds performance tags. Nano is an option to benchmark on CPU.

Explore Chatterbox ↗
VOICE DIRECTION

Qwen3-TTS

Apache 2.0. The 1.7B models offer natural-language voice direction and design. A bigger experiment with higher hardware demands.

Explore Qwen3-TTS ↗