Use Azure OpenAI with Interviewdom
Create an Azure OpenAI resource, deploy a model, copy the endpoint and key, and connect them to Interviewdom.
- Updated November 19, 2026
- Advanced
- About 20 min
- Payment card needed
- Can also listen (speech-to-text)
1. At a glance
- What you get: answers and screenshot reading (with a vision-capable deployment), plus speech-to-text with a transcription deployment.
- Good for: people whose company already uses Microsoft Azure.
- Cost: pay as you go on your Azure bill (see section 8).
- Time needed: about 20 minutes.
- Words to know: a deployment is your named copy of a model. Interviewdom calls the deployment name, not the model name.
2. Create an account
- You need an Azure subscription. Create one for free if you don't have one.
- Your account needs permission to create Azure OpenAI resources and deploy models.
3. Billing and a budget
Azure OpenAI is billed on your Azure subscription.
Set a budget alert (recommended):
- In the Azure portal go to Subscriptions, select your subscription, then Budgets.
- Select Add, enter a monthly amount and add an email alert (for example at 80%).
Microsoft is clear that budgets only notify you: "Resources aren't affected, and your consumption isn't stopped." Cost data can take 8–24 hours to show up.
4. Create the resource, deploy a model and get the key
Create the resource
- In the Azure portal, select Create a resource and search for Azure OpenAI. Select Create.
- Fill in Subscription, Resource group, Region, Name (for example
interviewdom-openai) and Pricing Tier (Standard). - Keep network access as All networks unless your company requires otherwise.
- Select Next until Review + submit, then Create. When done, select Go to resource.
Deploy a model
- Sign in to Microsoft Foundry (these steps use Foundry (classic); make sure the New Foundry toggle is off).
- Find and select your resource. If offered an upgrade to Foundry, you can select Cancel.
- Select Deployments (or Models + endpoints if you upgraded).
- Select + Deploy model → Deploy base model, pick a model and Confirm.
- Choose a Deployment name (for example
interview-chat) and select Deploy. Write the name down. - For listening, deploy a transcription model too (Microsoft recommends gpt-transcribe) and note that deployment name.
Copy the endpoint and key
- Back on your resource in the Azure portal, open Keys and Endpoint.
- Copy KEY1 and the Endpoint (it looks like
https://YOUR-RESOURCE.openai.azure.com/). Azure gives two keys so you can rotate one without downtime.
5. Pick a model
Which models you can deploy depends on your Region. For live answers, deploy a fast model with image input.
6. Connect in Interviewdom
- Go to Settings → AI Providers → Azure OpenAI.
- Paste the Endpoint, your chat Deployment name and KEY1.
- Leave API version empty for chat. Interviewdom uses Azure's v1 API (
…/openai/v1/), which doesn't need one. - Click Test connection. You should see something like OK · 450 ms.
Your key is stored encrypted on your PC and is sent only to your Azure resource, never to Interviewdom.
7. Speech-to-text
Azure OpenAI can transcribe with a transcription deployment:
- In Settings → AI Providers → Speech-to-text, choose Azure OpenAI.
- Enter the transcription deployment name and, if asked, API version 2024-02-01 (the version used in Microsoft's speech-to-text quickstart).
- Files are limited to 25 MB, which is fine for live chunks.
8. Costs in practice (estimate)
Cost depends on the model you deploy; see Azure OpenAI pricing. Assumptions for a 1-hour interview: about 80,000 input + 12,000 output tokens and 60 minutes of audio.
9. Troubleshooting
| Problem | What it means | Fix |
|---|---|---|
| 401 access denied | Wrong key or wrong endpoint | Copy KEY1 and the Endpoint again from Keys and Endpoint. |
| 404 deployment not found | You typed the model name instead of the deployment name, or a typo | Use the exact deployment name from Foundry. |
| 429 rate limit | The deployment's tokens-per-minute quota is used up | Wait, or raise the deployment's TPM in Quotas. |
| Request blocked by content filter | A content filter is assigned to each deployment | Review the deployment's Content Filter setting. |
| Network error | Your resource only allows selected networks, or a proxy blocks it | Check the resource's Networking settings or ask IT. |
10. Security tips
- Never share keys, including with Interviewdom support.
- Rotate keys: switch Interviewdom to KEY2, then regenerate KEY1.
- Interviewdom stores your key encrypted on your PC and sends it only to your Azure resource.