Client events for Qwen-ASR-Realtime - Alibaba Cloud Model Studio

For a feature overview and complete sample code, see Real-time speech recognition - Qwen. For server-to-client events, see Server events for Qwen-ASR-Realtime.

Event lifecycle

A typical session follows this sequence:

Establish a WebSocket connection.
Send session.update to configure audio format, language, and VAD settings.
Send input_audio_buffer.append repeatedly to stream audio data.
In Manual mode, send input_audio_buffer.commit to trigger recognition for a complete utterance. In VAD mode, the server triggers recognition automatically.
Send session.finish to end the session, then disconnect after receiving the session.finished response.

session.update

Configures the session. Send this immediately after establishing the WebSocket connection to set the audio format, the language, and VAD parameters. If omitted, defaults apply. The server responds with a session.updated event on success.

Parameters

Parameter	Type	Required	Description
type	string	Yes	Fixed value: `session.update`.
event_id	string	Yes	A unique event ID.
session	object	Yes	Session configuration object. See the session configuration table below.

Session configuration

Parameter	Type	Required	Description
input_audio_format	string	No	Audio encoding format. Valid values: `pcm`, `opus`. Default: `pcm`.
sample_rate	integer	No	Audio sampling rate in Hz. Valid values: `16000`, `8000`. Default: `16000`. Setting `8000` causes server-side upsampling to 16,000 Hz (minor delay). Use `8000` only for natively 8,000 Hz audio like telephony.
input_audio_transcription	object	No	Speech recognition settings.
input_audio_transcription.language	string	No	Language of the audio. See the supported languages table below.
input_audio_transcription.corpus.text	string	No	Context text for contextual biasing -- background text, entity vocabularies, or reference material that improves recognition accuracy. Maximum: 10,000 tokens.
turn_detection	object	No	VAD configuration. Set to `null` for Manual mode. If present, VAD mode is enabled.
turn_detection.type	string	Required when `turn_detection` is set	Fixed value: `server_vad`.
turn_detection.threshold	float	No	VAD sensitivity threshold. Default: `0.2`. Valid range: `[-1, 1]`. Lower values increase sensitivity (may trigger on background noise). Higher values reduce sensitivity and avoid false triggers in noisy environments. See recommended VAD presets below.
turn_detection.silence_duration_ms	integer	No	Silence duration in milliseconds marking utterance end. Default: `800`. Valid range: `[200, 6000]`. Shorter durations (e.g., 300 ms) speed up responses but may split natural pauses. Longer durations (e.g., 1,200 ms) handle pauses better but increase latency. See recommended VAD presets below.

Recommended VAD presets

Use these presets as starting points. Adjust based on your results:

Preset	threshold	silence_duration_ms	Best for
Low latency	`0.0`	`400`	Fast-paced interactions like voice commands or agent assist, where quick responses matter more than handling long pauses
Balanced (default)	`0.2`	`800`	General-purpose transcription with a balance between responsiveness and accuracy

Supported languages

Code	Language
zh	Chinese (Mandarin, Sichuanese, Minnan, and Wu)
yue	Cantonese
en	English
ja	Japanese
de	German
ko	Korean
ru	Russian
fr	French
pt	Portuguese
ar	Arabic
it	Italian
es	Spanish
hi	Hindi
id	Indonesian
th	Thai
tr	Turkish
uk	Ukrainian
vi	Vietnamese
cs	Czech
da	Danish
fil	Filipino
fi	Finnish
is	Icelandic
ms	Malay
no	Norwegian
pl	Polish
sv	Swedish

Example

{
    "event_id": "event_123",
    "type": "session.update",
    "session": {
        "input_audio_format": "pcm",
        "sample_rate": 16000,
        "input_audio_transcription": {
            "language": "zh"
        },
        "turn_detection": {
            "type": "server_vad",
            "threshold": 0.0,
            "silence_duration_ms": 400
        }
    }
}

input_audio_buffer.append

Streams an audio chunk to the server's input buffer -- the core event for sending audio data. Behavior differs by interaction mode:

VAD mode: The server monitors the buffer for voice activity and automatically triggers recognition.
Manual mode: The client controls utterance boundaries. Send smaller chunks for lower latency.

The audio field contains Base64-encoded data. In Manual mode, maximum size per event: 15 MiB. The server does not send a confirmation response.

Parameters

Parameter	Type	Required	Description
type	string	Yes	Fixed value: `input_audio_buffer.append`.
event_id	string	Yes	A unique event ID.
audio	string	Yes	Base64-encoded audio data.

Example

{
    "event_id": "event_2728",
    "type": "input_audio_buffer.append",
    "audio": "<Base64-encoded-audio-data>"
}

input_audio_buffer.commit

Triggers recognition for all audio in the buffer as a single utterance. Use in Manual mode when your application controls utterance boundaries (e.g., push-to-talk). Disabled in VAD mode. The server responds with an input_audio_buffer.committed event on success.

Parameters

Parameter	Type	Required	Description
type	string	Yes	Fixed value: `input_audio_buffer.commit`.
event_id	string	Yes	A unique event ID.

Example

{
    "event_id": "event_789",
    "type": "input_audio_buffer.commit"
}

session.finish

Ends the session. Server response depends on speech detection:

Speech detected: The server completes final recognition, sends a conversation.item.input_audio_transcription.completed event with the result, then sends a session.finished event.
No speech detected: The server sends session.finished directly.

After receiving session.finished, disconnect the WebSocket connection.

Parameters

Parameter	Type	Required	Description
type	string	Yes	Fixed value: `session.finish`.
event_id	string	Yes	A unique event ID.

Example

{
    "event_id": "event_341",
    "type": "session.finish"
}

​Event lifecycle

​session.update

​Parameters

​Session configuration

​Recommended VAD presets

​Supported languages

​Example

​input_audio_buffer.append

​Parameters

​Example

​input_audio_buffer.commit

​Parameters

​Example

​session.finish

​Parameters

​Example

Event lifecycle

session.update

Parameters

Session configuration

Recommended VAD presets

Supported languages

Example

input_audio_buffer.append

Parameters

Example

input_audio_buffer.commit

Parameters

Example

session.finish

Parameters

Example