Volcengine STT

Transcribe audio to text using Volcano Engine (Volcengine/ARK) speech-to-text APIs. Use when the user wants to replace Whisper/OpenAI STT with Volcengine, transcribe Telegram/Discord voice notes via Volcengine, or build a reusable STT skill for other OpenClaw agents.

Use this skill to run speech-to-text through Volcengine.

Quick start

{baseDir}/scripts/transcribe.sh /path/to/audio.ogg

Default behavior:

• Endpoint: ${ARK_BASE_URL:-https://ark.cn-beijing.volces.com/api/v3}/audio/transcriptions
• Model: ${ARK_STT_MODEL:-doubao-seed-asr-1-0}
• Auth header: Authorization: Bearer $ARK_API_KEY
• Output file: <input>.txt

Required env

• ARK_API_KEY (required)

Optional:

• ARK_BASE_URL (default: https://ark.cn-beijing.volces.com/api/v3)
• ARK_STT_MODEL (default: doubao-seed-asr-1-0)

Useful flags

# Save plain text to custom path
{baseDir}/scripts/transcribe.sh ./voice.ogg --out /tmp/voice.txt

# Force model
{baseDir}/scripts/transcribe.sh ./voice.ogg --model doubao-seed-asr-1-0

# Return raw JSON (for debugging/integration)
{baseDir}/scripts/transcribe.sh ./voice.ogg --json --out /tmp/voice.json

# Hint language/prompt when needed
{baseDir}/scripts/transcribe.sh ./voice.ogg --language zh --prompt "中英混合，保留术语"

Integration notes

• For OpenClaw voice-message handling, call this script instead of Whisper script.
• Keep keys in machine-local config or env, never commit secrets.
• If your Volcengine account uses a different model name, pass --model or set ARK_STT_MODEL.