Aliyun Speech Transcriber

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.

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:

powershell

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

Multiple files:

powershell

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:

1. Use qiniu-upload to upload a local file.
2. Prefer a signed private URL if the domain is not anonymously readable.
3. 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.