Aliyun Speech Transcriber

该 Skill 调用 Node 脚本，将用户提供的可公开访问音频或视频 URL 提交至阿里云 DashScope 的 paraformer-v2 文件转写服务，轮询任务并返回包含任务标识、结果和尽力提取文本的 JSON。需要配置 DashScope API Key，支持一次传入多个 URL；不包含本地文件上传或七牛上传流程。

chenggongdu

热度 52.32

最近更新 2026-05-11 17:35:38

分类：效率依赖：代码执行 / 第三方 API应用场景：数据处理 / 文件转换

安装技能

复制下面的话术发给你的 Agent，它会通过 secure-skills 检索并安装该技能。

执行命令安装此 skill： npx secure-skills-cli add aliyun-speech-transcriber

Metadata

name: aliyun-speech-transcriber
description: Transcribe publicly accessible audio or video URLs with Aliyun speech services. Use when the user wants speech-to-text via Aliyun DashScope, needs transcript JSON or extracted plain text, or wants to process a cloud-accessible media URL (including signed Qiniu URLs) into transcription results.
homepage: https://dashscope.aliyun.com/
metadata: {"clawdbot":{"emoji":"🎤","requires":{"env":["ASR_DASHSCOPE_API_KEY"]},"tags":["speech","transcription","asr","audio","aliyun","dashscope"]}}

Aliyun Speech Transcriber

Use this skill to turn externally accessible media URLs into transcript results.

Current scope

Current implementation focuses on DashScope file transcription using the paraformer-v2 model, aligned with the existing Java service pattern.

Required environment variables

ASR_DASHSCOPE_API_KEY

Fallback supported:

DASHSCOPE_API_KEY

Optional:

ALIYUN_SPEECH_MODEL - defaults to paraformer-v2
ALIYUN_SPEECH_LANG_HINTS - defaults to zh,en
ALIYUN_SPEECH_POLL_SECONDS - defaults to 5
ALIYUN_SPEECH_TIMEOUT_SECONDS - defaults to 1800

Inputs

Pass one or more externally accessible URLs:

node scripts/transcribe.js --file-url "https://example.com/audio.mp3"

Multiple files:

node scripts/transcribe.js --file-url "https://a.com/1.mp3" --file-url "https://a.com/2.mp3"

Output

The script returns JSON with:

success
provider
engine
taskId
requestId
results
text

text is a best-effort plain-text extraction from the final JSON result.

Chaining from Qiniu

Typical workflow:

Use qiniu-upload to upload a local file.
Prefer a signed private URL if the domain is not anonymously readable.
Pass the returned URL into this skill.

Safety rules

Never hardcode Aliyun credentials.
Fail fast if DASHSCOPE_API_KEY is missing.
Only send URLs the user intends to transcribe.