Azure OpenAI Whisper Service

One of the great things about my role is that I get to work with really great, really interested technical evangelists. One of the things we got chatting about today was the new Whisper service and how I could upload text into Azure and have the new OpenAI Whisper service translate it for me and transcribe it. If anyone knows me, they know I don't like typing. My brain and my hands don't always work at the same speeds. And anyone who's read any of my Team’s messages, they can tell sometimes the message is garbled and quite often the words are in the wrong order. So being able to talk and have that transcribed is a big thing for me and it's something that I feel will be useful. So today I've been testing this service and I'm recording this as we speak. I'm just using a Microsoft sound recorder. I'm just capturing it as a file, recording it on my machine. I'll then upload it into the Azure Whisper service through the Microsoft speech portal and we'll get it transcribed.
So, let's see the results.


So that took at lot less time to record and edit than it would to type. 93 seconds to record, 5 seconds to upload and a minute to edit.

I hadn’t used Speech studio before, it really couldn’t be easier, so now I have no excuse for not blogging more.

I have tried dictate in Word it's not so good and I get fustrated but this is well cool.

Components:
  • Azure OpenAI
  • Whisper Model (West Europe and North Central US)
  • Sound Recorder
  • Azure Speech Studio
Azure OpenAI
Step 1: Create an OpenAI instance in West Europe or North Central

Step 2: Use the Azure AI Studio to deploy the whisper model.

Open AI Whisper Model


Step 3: Head over to the Microsoft Speech Studio

Microsoft Speech Studio


Step 4: Select to OpenAI Whisper Service

Microsoft Speech Studio - OpenAI Whisper


Step 5: Record the Audio

Windows Sound Recorder


Step 5: Select the OpenAI instance

Select the OpenAI instance


Step 6: Try it out

Speech to Text