Overview
Fun-ASR-Realtime is an end-to-end, real-time speech recognition model for converting live audio into text. Alibaba lists this version in its September 23, 2025 release history. (docs.modelstudio.console.alibabacloud.com)
Capabilities and uses
The release announcement describes contextual recognition, customizable hotwords, inverse text normalization, punctuation prediction, filtering of sensitive words and filler words, and noise robustness. The version-specific model documentation lists Mandarin Chinese and English, WebSocket access, and hotword support. Suitable applications include live captions, meeting transcription, and voice input. (docs.modelstudio.console.alibabacloud.com)
Limitations and availability
The version-specific documentation lists emotion recognition and speaker diarization as unsupported. Alibaba announced retirement of this model effective October 10, 2026, subject to the actual change time, and recommends fun-asr-realtime as a replacement. Actual availability was not independently tested. (help.aliyun.com)
Pricing
Alibaba Cloud's retirement notice lists a historical input price of USD 0.000047 per second for this exact model and states that its listed prices are based on the Singapore region. The current model-pricing documentation describes real-time speech recognition as billed by input-audio duration, with output not billed. This historical price should not be assumed to indicate current availability. (alibabacloud.com)
The current domestic pricing page does not list this exact snapshot, so its applicable CNY price could not be verified. Prices were checked on November 18, 2026. (help.aliyun.com)