OpenRouter / Speech to text

Transcribes a short audio clip with one of 24 speech-to-text models on OpenRouter: OpenAI whisper and gpt-4o-transcribe, Deepgram Nova-3, AssemblyAI Universal, ElevenLabs Scribe v2, Google Chirp 3 and Gemini transcribe, Mistral Voxtral, Qwen3 ASR, NVIDIA Parakeet, Fish Audio, Grok STT. The audio goes inline as base64 in input_audio.data with its format (mp3, wav, m4a, ogg, flac, webm), up to about 700 KB of audio because a looot request is capped at 1 MB. Returns text and usage with the exact cost; response_format verbose_json adds segments and timestamps on OpenAI-style models. Charged at OpenRouter's reported cost plus its 5.5% credit fee.