Developers building voice and audio features may like the cheap TTS, but the Realtime, Gen and Music models are still free-for-now previews.

Kenning's take

What it does

StepAudio 3 is a family of five models from StepFun: Realtime for live voice conversations, ASR for transcription, TTS for speech, Gen for mixed speech, sound effects and background music, and Music for songs. Developers use them through the StepFun API, and they can be tried in StepFun's Voice Studio. The Realtime, Gen and Music models are previews that are free for a limited time.

• Full-duplex Realtime voice model with tool calls
• ASR for names, jargon, dialects and code-switching
• TTS with laughter, hesitation and self-correction
• Gen mixes speech, sound effects, ambience and music
• Music generation from text, lyrics and audio

Who it suits. Developers building voice agents, transcription services or audio and music generation into apps.

Key facts

Category

AI models

Made by

StepFun

Based in

China

Released

15 September 2026

How to get it

through the StepFun API; try in Voice Studio at audio.stepfun.ai

Works on

API, Web (Voice Studio)

Free plan

stepaudio-3-realtime-preview, stepaudio-3-gen-preview and stepaudio-3-music-preview are free for a limited time

API price

stepaudio-3-tts $0.36 per 10,000 characters; stepaudio-3-asr-max $0.40 per hour; voice cloning $1.50 per voice

Reviews and write-ups

• ifeng Tech 2026-09-15

FAQ

Facts from the maker's own pages and the sources listed. Checked 2026-10-11.