asr-transcribe

name	asr-transcribe
description	Use when a user sends a voice message, uploads an audio file, or provides a public audio URL and wants a text transcription.

ASR Transcribe

Overview

Transcribe speech to text with DashScope native ASR using qwen3-asr-flash.

When to Use

• User sends a voice message and wants the spoken content as text.
• User uploads an audio file and explicitly asks for transcription.
• User provides a public audio URL and asks for transcription.

Do not use this skill for TTS or for ordinary text-only requests.

Inputs

• Absolute local audio file path, or
• Public http:// or https:// audio URL

Configuration

• OPENCLAW_DASHSCOPE_ASR_API_KEY
• OPENCLAW_DASHSCOPE_API_KEY fallback
• OPENCLAW_DASHSCOPE_ASR_BASE_URL optional override
• OPENCLAW_DASHSCOPE_ASR_MODEL optional override, defaults to qwen3-asr-flash

Use

1. Resolve one audio source.
2. Run:
   • node transcribe.mjs --input /absolute/path/to/audio.wav
   • or node transcribe.mjs --input https://example.com/audio.mp3
3. Return the transcript text to the user.
4. If the helper prints JSON with --json, extract the text field unless the user asked for metadata.

Notes

• Prefer the saved local file path for channel voice messages.
• Keep the reply focused on the transcript unless the user asks for timestamps or raw metadata.