StepAudio 3 ASR Max
StepAudio 3 ASR Max combines acoustic modeling with language-model context. The model guide covers Chinese, English, dialects, mixed-language speech and specialist terms.
Use context to resolve similar sounds
The model is designed to use domain knowledge and surrounding meaning for names and technical words. This differs from a general audio conversation model that produces a free-form reply.StepFun stepaudio-3-asr guide
Review critical names in the result
Practical guidance. Context can improve a transcription, but it can also make an incorrect word look plausible. Compare proper names, numbers and specialist terms with the original recording. Use a representative set of your domain's audio for evaluation.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | API, Provider product |
| Model identifiers | stepaudio-3-asr-max |
| Interface | The current SSE API recommends stepaudio-3-asr-max.StepFun speech recognition with SSE |
| Output | Incremental and final text over HTTP with server-sent events.StepFun stepaudio-3-asr guide |
| Audio tasks | The guide includes whispered speech, rapid speech and sung lyrics.StepFun stepaudio-3-asr guide |
Related models
- StepAudio 2 ASR Pro
StepFun · StepAudio 2
StepAudio 2 ASR Pro is the 32B recognition model in StepFun's audio catalog.
Launch date not confirmedAPI - StepAudio 2.5 ASR
StepFun · StepAudio 2.5
StepAudio 2.5 ASR is a 4B speech recognition model that uses multi-token prediction.
Launch date not confirmedAPI