🎙️ Voice Clone Bench — Chatterbox (zero-shot)

Upload a reference voice, type any text, get speech in that cloned voice. Built to A/B against ElevenLabs. Model: Chatterbox Multilingual (Resemble AI, MIT).

① Reference voice to clone (5–20s clean speech). Empty = built-in sample.

0:00

Strips music/noise with HT-Demucs so the clone is built from clean speech.

🧹 Remove background audio from reference (isolate voice before cloning)

Isolated voice (preview)

② Language

③ Text to speak (long scripts are auto-chunked)

Exaggeration — lower = more neutral/faithful (≈0.4); 0.7+ = expressive

0 2

CFG / Pace — 0.5 balanced; ~0.3 faster; 0.0 for cross-lingual

0 1

Temperature — lower = more consistent/faithful (≈0.7)

0.05 2

Seed (0 = random)

Cloned speech output