xby-asr-zh

If your DSH workflow keeps pulling you out of the conversation to transcribe a Chinese audio clip into text, xby-asr-zh closes that loop. One command inside the chat takes audio in and returns clean text, no tab-switching to an external tool required.

Under the hood it runs the Paraformer INT8-quantised model, tuned specifically for Chinese speech, delivering fast inference with a small memory footprint. The plugin accepts audio in three shapes — a plain URL, a Base64-encoded payload, or a local file path — so whatever the audio is sitting in right now, you can hand it straight to the model without a conversion step. The API key you set during the chat is persisted automatically, meaning a restart does not wipe your configuration.

It is a natural fit when you are building voice-assistant pipelines, transcribing meeting recordings, processing podcast episodes, or simply need to turn a Chinese voice memo into searchable text without leaving the DSH environment. Once the plugin is installed and the key is in place, every subsequent recognition call is a single function invocation.