Breeze TTS 2
playground
checking model…
API key
— stored only in this browser
Show
Prompt
— describe the voice & delivery (optional)
Calm woman
Narrator
Radio host
Farmer
Keep it short and specific (under ~1000 characters). Write it in the same language as the script. Leave empty for the default voice.
Script
— the text to speak
0 / 5000
English & Chinese. Inline events:
(laugh)
(cough)
(sigh)
· 中文:
[笑]
[叹气]
Advanced · reference voice (clone / direction), CFG, seed
CFG scale
(blank = auto: 4 with prompt, else 1)
Seed
Reference audio (clean speech, ≤ 20 MB)
Exact transcript of the reference audio
Reference only = voice clone · reference + prompt = voice direction.
Generate audio
Result
⬇ Download WAV
This session
Use it as an API
Long scripts (up to 5000 chars) are split automatically at sentence boundaries and stitched together. Returns
audio/wav
(24 kHz, mono, 16-bit). Send the key as
Authorization: Bearer …
or
X-API-Key
. Requests are queued: one generation runs at a time.