What problem does it solve? Converting audio or video recordings into text requires calling a speech-to-text API with the right model, format, and provider settings, which is tedious to script by hand. This Skill wraps the OpenAI Audio Transcriptions endpoint in a single curl-based command that handles provider detection, output formatting, and error checking. ## Core Features & Use Cases - Audio/Video Transcription: Send any audio or video file (m4a, ogg, mp4, etc.) to the /v1/audio/transcriptions endpoint and receive plain text, JSON, or verbose JSON output. - Subtitle Generation: Produce SRT or VTT caption files with timestamps by requesting verbose JSON and assembling subtitle cues locally, which works even against providers like Groq that do not emit SRT natively. - Provider-Aware Defaults: Automatically selects whisper-large-v3-turbo for Groq/vveai endpoints and whisper-1 otherwise, with overrides via --model or WHISPER_MODEL. - Use Case: Given a recorded interview saved as interview.m4a, run the transcribe script with --srt to produce timestamped captions ready for a video editor. ## Quick Start Transcribe the attached audio file into text using the Whisper transcription script, optionally adding --srt to get timestamped subtitles.
