voice-transcription

name	voice-transcription
description	Transcribe audio files (e.g. Discord .ogg voice messages) using the OpenAI Whisper API. Use when: a voice message or audio attachment needs to be converted to text. NOT for: real-time streaming audio, files >25MB.
homepage	https://platform.openai.com/docs/guides/speech-to-text
metadata	{"openclaw":{"emoji":"🎙️","requires":{"bins":["curl","jq"],"env":["OPENAI_API_KEY"]}}}

Voice Transcription Skill

Transcribe audio files to text using the OpenAI Whisper API.

When to Use

USE this skill when:

• A voice message (e.g. Discord .ogg) needs to be read as text
• An audio recording needs to be converted to text
• A meeting or interview recording needs a transcript
• Any supported audio file needs speech-to-text conversion

When NOT to Use

DON'T use this skill when:

• Real-time or streaming audio transcription is needed
• Audio file is larger than 25MB (Whisper API limit)
• No OPENAI_API_KEY is available
• Format is not in the supported list below

Command Syntax

bash skills/voice-transcription/transcribe.sh <audio-file-path>

The transcription is printed to stdout. All errors go to stderr.

Supported Formats

flac, mp3, mp4, mpeg, mpga, m4a, ogg, opus, wav, webm

Cost

Whisper API pricing: $0.006 per minute of audio.

Every call is logged to $HOME/hivewright/logs/voice-transcription-cost.log.

Error Codes

Exit Code	Cause
1	No file path argument provided
1	File not found at given path
1	Unsupported file format
1	curl or jq not installed
1	OPENAI_API_KEY not set
1	Whisper API returned non-200 response
1	Empty or null transcription returned

Log Location

$HOME/hivewright/logs/voice-transcription-cost.log

Each line: YYYY-MM-DD HH:MM:SS | file=<basename> | size=<bytes>B | model=whisper-1 | cost=~$0.001-0.006