OpenAI Makes Whisper Speech Recognition Available on OpenRouter at $0.006 per Minute

OpenAI's Whisper 1 automatic speech recognition model is now accessible through OpenRouter's API routing service, priced at $0.006 per minute of audio duration.

Technical Specifications

Whisper 1 accepts audio files up to 25 MB in multiple formats including mp3, mp4, wav, and webm. The model supports transcription and translation across more than 50 languages. Billing is calculated per minute of audio duration, rounded to the nearest second.

The model is OpenAI's open-source speech recognition system, previously available through OpenAI's direct API as whisper-1. OpenRouter now provides routing to multiple providers hosting the model, with automatic fallbacks to maximize uptime.

API Access

OpenRouter normalizes requests and responses across providers through an OpenAI-compatible completion API. According to the routing service, it automatically selects the best available provider based on prompt size and parameters.

Developers can access Whisper 1 using the OpenAI SDK, Anthropic SDK, or OpenRouter's native SDK. The service supports standard audio transcription and translation workflows without provider-specific code changes.

Pricing Model

At $0.006 per minute, Whisper 1 on OpenRouter charges for actual audio duration rather than token-based pricing used by text models. A 10-minute audio file costs $0.06 to process. The pricing structure remains consistent regardless of which underlying provider OpenRouter routes to.

What This Means

OpenRouter's addition of Whisper 1 extends its model catalog beyond text and multimodal models into speech processing. The routing layer allows developers to integrate speech recognition without managing multiple provider relationships or handling failover logic. However, OpenRouter has not disclosed usage statistics or uptimes for the newly added model. The $0.006 per minute pricing matches OpenAI's direct API pricing, meaning developers gain routing flexibility without cost penalty.