Azure Speech-to-Text: language or quota limitaion

Question

I want to use Azure's speech-to-text for my chatbot service.

The use-case is to provide STT results when the user gives voice (usually one or two sentences). Here, the user does not select a language.

I tried using Azure OpenAI's whisper, but the quota was too small (30 requests per minute even after increasing it).

For language identification in Azure speech AI, I saw in the documentation that candidate languages ​​are limited to 4 (10 for consecutive LIDs).

So, my question is:

1. I previously increased the quota of azure openai whisper from 3 RPM(requests per minute) to 30. Is it possible to increase this infinitely (or almost infinitely)?
2. In the case of Azure speech AI's language identification, is there a way to search for a language for all languages ​​provided by Azure without being limited to candidate languages ​​(or limited to at least 20 candidate languages)?

Answer 1

Hi 이찬규!

Welcome to Azure AI Q and A forum. Thank you for posting your query here.

Here are the answers to your queries.

1. Limitation on Whisper model RPM: Quota allotment for whisper and other OpenAI models happens only based on regional availability and use case of customer and cannot be increased indefinitely to provide fair distribution to all customers based on use case, existing quota usage, expected traffic and demand of customer. Reference Customers need to apply by sharing their details and justification on quota increase form.
2. Limits on language identification: You can include up to 4 languages for at-start LID or up to 10 languages for continuous LID. Kindly refer this section on candidate language.

If this answers your query, do click Accept Answer and Yes for was this answer helpful. And, if you have any further query do let us know.

Thank you.