Qwen Alibaba Cloud: Qwen3 ASR Flash
Qwen multilingual speech recognition model for pre-recorded audio, covering 27 language codes with automatic language detection, context biasing, optional inverse text normalization and an always-on emotion label.
バリアントモデル
同じ正規モデルファミリーに対して、作成者やプロバイダーが受け付けるネイティブのモデルIDです。
Stable id, currently equivalent to the qwen3-asr-flash-2025-09-08 snapshot rather than the latest one.
Latest dated snapshot of Qwen3 ASR Flash, listed separately from the stable id.
Earlier dated snapshot of Qwen3 ASR Flash; the stable qwen3-asr-flash id currently resolves to this snapshot.
プロバイダー
同じモデルを複数のプロバイダーがホストしている場合があります。バックエンドを固定したい場合は1つのプロバイダーを選び、そうでなければルーティングに選択を任せられます。
対応言語
zh
yue
en
ja
de
ko
ru
fr
pt
ar
it
es
hi
id
th
tr
uk
vi
cs
da
fil
fi
is
ms
no
pl
sv
機能
Transcribes spoken audio to text.
speech_to_text
Transcribes a fully-submitted audio file in a single request.
synchronous
Detects the spoken language automatically when no language is specified and reports it with the transcript.
language_identification
Supports 27 language codes, with Mandarin, Sichuanese, Hokkien/Minnan and Wu all under zh and Cantonese under the separate yue code.
multilingual_transcription
Accepts free-form context text (background text or hotwords) that biases recognition toward domain-specific terms.
context_biasing
Returns an always-on emotion label (neutral, happy, sad, angry, disgusted, fearful, surprised) with the transcript.
emotion_recognition
Optional inverse text normalization, off by default and effective for Chinese and English only.
inverse_text_normalization
コード
ElevenLabs SDKでのストリーミングSTT
マイクまたはファイルの音声を、ElevenLabs互換のAllModelsルート経由でストリーミングします。
import { ElevenLabsClient } from "elevenlabs";
const client = new ElevenLabsClient({
apiKey: process.env.ALLMODELS_API_KEY,
baseUrl: "https://api.allmodels.io/el"
});
const connection = await client.speechToText.realtime.connect({
modelId: "qwen/qwen3-asr-flash",
languageCode: "en"
}, {
queryParams: {
provider: {
"only": [
"qwen"
]
}
}
});
// Send audio frames to connection, then handle transcript events.