An Azure service that integrates speech processing into apps and services.
1. Access to LLM Speech Transcription
- LLM Speech is available as part of Azure Speech in Microsoft Foundry. It enhances speech recognition with a large language model, providing improved transcription quality, multilingual support, contextual understanding, translation, and prompt-based customization. To use LLM Speech:
- You need an active Azure subscription.
- You need a Microsoft Foundry project and an Azure Speech resource.
- Your application can access the service through the supported REST APIs and SDKs when the resource is deployed in a supported region.
- Current service limits include:
- Maximum audio file size: 500 MB
- Maximum audio length: 5 hours per file
- Up to 600 requests per minute per Speech resource.
- Speech resource region
- Speech resource SKU (F0 or S0)
- Any API error message or response code
2. Pricing and Billing
According to the Azure Speech pricing documentation, traditional Speech-to-Text services are billed based on audio duration (per-second billing). Standard real-time transcription and fast transcription have different pricing tiers. For LLM Speech, Microsoft documentation currently provides feature availability and service limits but does not explicitly publish a separate LLM Speech pricing meter or rate on the referenced pricing and documentation pages. Therefore:- Standard Speech-to-Text pricing is publicly documented and billed by audio duration.
- LLM Speech feature documentation does not currently specify a dedicated billing unit or rate.
- If your organization requires definitive billing details for LLM Speech, we recommend checking the Azure Pricing Calculator for your subscription or opening an Azure billing support request, as pricing can vary by region and offer type.
Thanks