What's New

StepAudio 3 TTS

OFFICIAL OVERVIEW

As natural as a real person, and more than speech

StepAudio 3 TTS delivers industry-leading TTS and voice design, bringing lifelike expression and flexible voice control to zero-shot speech generation

Built on shared RVQ tokens with a discrete autoregressive architecture, it can also generate singing, sound effects, music, and their blends — all unified in a single model

StepAudio 3 ASR

Hear every word. Understand the field.

The next stage of speech recognition is not just catching every word, but truly understanding what the speaker means — across dialects, noise, technical jargon, and contextual ambiguity. StepAudio 3 ASR Max pairs high-precision acoustic modeling with a large language model's contextual understanding, knowledge, and reasoning, moving recognition from “hearing sound” to “understanding context.”

StepAudio 3 Realtime

Natural conversation should feel effortless

In everyday conversation, it picks up naturally and responds with a relaxed, human cadence. When the exchange gets lively, you can chime in, add a thought, or interrupt whenever you want. Extremely low latency and smooth full-duplex interaction make every pause, response, and turn feel fluid and natural, more like talking with another person.

StepAudio 3 Gen

Vocals that move you.

Describe the singing style, voice, and emotion to generate a cappella vocals.

The Warmth of Goodbye告别余温

A restrained, heartfelt male voice that swells into the chorus, carrying the lingering warmth of goodbye.克制而深情的男声,副歌渐强,唱出告别时的余温与不舍。

Lyrics: 歌词:再见说了千百遍,这次终于是真的。你转过的那个弯,把我的春天也带走了。

Pop balladMaleChinese抒情流行男声中文
Coconut Afternoon椰风午后

A lazy, relaxed Bossa Nova vocal over nylon guitar and shaker, bright as an afternoon in the sun.慵懒松弛的 Bossa Nova 女声,尼龙吉他与沙锤轻轻托着午后阳光。

Lyrics: 歌词:The sea breeze crinkles my skirt, you hum a tune off-key, an afternoon inside a coconut shell, slow like sugar that never melts.

Bossa NovaFemaleEnglishBossa Nova女声英文
Home Is Fine家书

A warm, storytelling voice reading out a letter written home, line by line.温暖的叙事女声,一字一句读出一封写给家的信。

Lyrics: 歌词:A few more white hairs on Mom. The phone always says home is fine.

Pop balladFemaleEnglish流行抒情女声英文
Be Well你要好好的

A soft slow ballad — breathy vocals wrap the chorus in a quiet blessing.轻柔的女声慢歌,气声包裹着祝福,副歌处轻轻扬起。

Lyrics: 歌词:我把再见咽下去,换成一句你要好好的。

Slow balladFemaleChinese抒情慢歌女声中文
Aria of Fury怒焰咏叹

A bright, powerful young male voice, full-breathed and high-charged with indignation — a D-major aria at a blazing allegro.明亮有力的青年男声,气息充沛,声线高亢激昂,带着愤怒控诉的情绪。D 大调咏叹调,快板,激昂澎湃。

Lyrics: 歌词:你怎能如此背叛我的信任!这火焰般的愤怒在我胸中燃烧。我要让全世界听见我的呐喊,让背叛者付出代价!

OperaMaleChinese歌剧男声中文
Neon Afterglow霓虹未眠

A neon-lit city-pop vocal over synth pads, glowing alone at night.霓虹质感的都市流行男声,合成器铺底,在夜里独自发亮。

Lyrics: 歌词:霓虹在窗台慢慢晕开,像你上次没说完的对白。地铁最后一班带走喧哗,留我数着红灯发呆。如果想念有形状,该是这城市忽明忽暗的光。我沿着晚风走回老地方,假装你还没走,还靠在我身旁。

PopMaleChinese流行男声中文

StepAudio 3 Music

Get the song clear in your head, then hear it sung.

StepAudio 3 Music pairs an MoE architecture with an AR + DiT generation paradigm and uses ABC-COT to plan a song’s structure, sections, and arrangement before it generates. It controls style, vocals, emotion, instrumentation, key, and tempo from your creative intent, so the result is a complete song that stays true to your description.

1. Your description

Style, vocals, timbre, emotion, instruments, key, tempo — plain language is enough

2. Structure & arrangement

Set the section flow and instrument roles first, then generate

3. Vocals & mixing

Vocals and backing follow the plan, with sections that flow naturally

4. A complete song

A finished piece you can play and download — not a fragment