Tasks and inputs
StepAudio 3 Realtime Preview is listed for Audio-text-to-text, Audio-to-audio. It accepts text, audio and produces text, audio. Use the publisher’s input format and preprocessing for your chosen deployment.
SyncDev model directory · Profile updated Oct 8, 2026
StepAudio 3 Realtime Preview combines spoken dialogue, turn handling, reasoning and tool calls over a persistent WebSocket.
The specification score summarizes recorded features and limits. It is not a benchmark result.
Recorded modality support shows what this model can accept and produce. An undocumented modality is shown as unknown rather than assumed to be unsupported.
Accepts
Produces
StepAudio 3 Realtime Preview combines spoken dialogue, turn handling, reasoning and tool calls over a persistent WebSocket.
The realtime route handles audio understanding and spoken responses in one conversation. Its documented features include voice activity detection, conversation history, interruption handling and tool use. It can continue speaking while a backend task runs, then incorporate the result when that task completes.
A useful voice assistant must distinguish an interruption from a short acknowledgement. Test a user who says “yes” while the assistant speaks, a user who changes the request halfway through and a conversation with background noise. Review whether the assistant stops appropriately and resumes with the correct task state. The publisher’s latency claims do not replace a measurement over your own network and audio devices.
Use stepaudio-3-realtime-preview on the Realtime API. The documentation describes temporary trial access and a transition to a paid version. Permanent audio-token prices are therefore left unknown here. Confirm the identifier and account access before connecting a production voice interface.
Give each tool a narrow purpose and a clear response format. Test failed calls and delayed results as well as successful ones, checking whether the conversation remains understandable while a task is outstanding. Keep a transcript and tool trace for diagnosing a misunderstood request or an incorrect action.
Compare the model on the work you need to complete.
StepAudio 3 Realtime Preview is listed for Audio-text-to-text, Audio-to-audio. It accepts text, audio and produces text, audio. Use the publisher’s input format and preprocessing for your chosen deployment.
Compare pronunciation, voice consistency and time to the first playable audio. Check the service’s billing unit for the actual script.
Use the model identifier and account requirements for the selected API route. Confirm its request limits and output format with one complete task before connecting a larger workflow.
These records share a directory category. Check the source card, output type and deployment requirements before treating a model as a replacement.
Alibaba
Qwen-Audio-3.1-Realtime-Plus supports duplex speech conversation, voice cloning, tools and a 262K-token context through its realtime interface.
Gemini 3.8 Live is Google's model for interactive audio conversations with text, image, audio and video input. Its Live API response modality is audio; text is available through transcription.
Aratako
Aratako/MioCodec-25Hz-24kHz is a source-linked Hugging Face repository assigned to Audio-to-audio. Its recorded task interface accepts audio and produces audio.
Aratako
Aratako/MioCodec-25Hz-44.1kHz-v2 is a source-linked Hugging Face repository assigned to Audio-to-audio. Its recorded task interface accepts audio and produces audio.
aufklarer
aufklarer/DeepFilterNet3-CoreML is a source-linked Hugging Face repository assigned to Audio-to-audio. Its recorded task interface accepts audio and produces audio.
aufklarer
aufklarer/PersonaPlex-7B-MLX-4bit is a source-linked Hugging Face repository assigned to Audio-to-audio. Its recorded task interface accepts audio and produces audio.