跳到正文
通義 QwenAudio:原創語音項目·· 2026-06-17AI 評分45

FunASR 接入 LlamaIndex Readers,支援本機語音轉文字

QwenAudio/llama-index-readers-funasr

AI 導讀

LlamaIndex Readers 的 FunASR 整合可將音訊轉錄為 LlamaIndex Document,使用 SenseVoice、Paraformer 或 Fun-ASR-Nano 在本機執行語音轉文字,毋須雲端 API。它擅長中文及 50+ 種語言,內置 VAD 可處理任何長度的音訊。

正文

LlamaIndex Readers Integration: FunASR

Transcribe audio into LlamaIndex Documents with FunASR — self-hosted speech-to-text powered by SenseVoice / Paraformer / Fun-ASR-Nano. Runs locally, no cloud API; strong on Chinese and 50+ languages.

Installation

pip install llama-index-readers-funasr

Usage

from llama_index.readers.funasr import FunASRReader

reader = FunASRReader(model="iic/SenseVoiceSmall", device="cuda")
documents = reader.load_data("meeting.wav")
print(documents[0].text)

Pass hub="hf" with model="FunAudioLLM/SenseVoiceSmall" to load from HuggingFace. The built-in VAD handles audio of any length.

來源:通義 QwenAudio:原創語音項目 · github.com