Transcribe audio to text using Whisper. Returns a job that you poll or stream for completion.
Supports both Fal-hosted Whisper and Groq-hosted Whisper (faster, cheaper).
API key from varg.ai/dashboard.
Pass as Authorization: Bearer varg_xxx.
Transcription model to use.
Fal (standard): whisper, whisper-large-v3
Groq (faster, cheaper): groq-whisper, groq-whisper-large-v3, groq-whisper-large-v3-turbo
"whisper"
"whisper-large-v3"
"groq-whisper"
"groq-whisper-large-v3-turbo"
URL of the audio file to transcribe (mp3, wav, m4a, webm, etc.)
"https://example.com/audio.mp3"
Language code (e.g. "en", "es", "fr", "ja"). Auto-detected if omitted.
"en"
Additional options passed directly to the provider
Job created
Unique job identifier
"job_a1b2c3d4"
Current job status
queued, processing, completed, failed, cancelled "completed"
Resolved model identifier (provider:model format)
"fal:kling-v2.5"
When the job was created
"2026-02-13T10:00:00Z"
When the job finished (completed, failed, or cancelled)
"2026-02-13T10:02:30Z"
Job output, present when status is completed
Cache metadata
Error message, present when status is failed
"Provider returned an error"