My new plan for tts voices is loading for too long

Welcome to Microsoft Q&A

Hello @Evonna Nash I hope you are doing well,

What is actually happening is that the transition to the "new look" of the Azure AI Foundry portal relies on entirely new frontend bundles and CDN routing paths. If you are experiencing unusually long loading times (or hanging on the voice catalog screen) compared to the old portal, it is typically caused by stale browser cache from the previous Azure AI Studio design, or a network firewall/VPN that is slowing down the new static asset domains required to render the interface.

Check please the following things:

1. Clear Specific Cache: Clear your browser cache and site data specifically for ai.azure.com. The new Foundry UI often struggles to load if it is trying to reconcile old UI cache data with the new layout engine.

Bypass Network Filters: Open an Incognito/InPrivate window and temporarily disconnect your VPN or ad-blocker. The new portal uses different backend endpoints to fetch the large TTS voice catalog, and local network filters inspecting these new URLs can cause severe latency.

Use the Classic Fallback: If the Foundry portal continues to load too slowly, you can completely bypass this UI issue by managing your Text-to-Speech voices and deployments directly in the dedicated Speech Studio portal, which retains the classic, fast-loading architecture.

References

• Cannot access AI Azure Foundry - Stuck on loading - Microsoft Q&A
• Azure AI services - Speech Studio

If my answer helped you resolve your issue, please consider marking it as the correct answer. This helps others in the community find solutions more easily.

Hello @Evonna Nash

Thank you for reaching out to Microsoft Q&A.

I understand that Text-to-Speech (TTS) voices are taking significantly longer to load in the new Azure AI Foundry experience, whereas the previous interface was working as expected.

However, there are a few areas that are worth checking, as they can sometimes present as prolonged loading times.

1. Verify whether the issue is limited to the Foundry UI

Please try the following basic troubleshooting steps:

• Open Azure AI Foundry in an InPrivate/Incognito browser window.

Test using a different browser (Microsoft Edge or Google Chrome recommended).

Clear browser cache and cookies and sign in again.

Check whether the behavior is reproducible across multiple devices or networks.

If speech synthesis works normally through APIs or other Speech tools but loads slowly only within Foundry, this would indicate a portal experience issue rather than a backend Speech service issue.

2. Check whether all voices are affected

Please let us know:

• Whether all voices are loading slowly or only specific voices.
• Whether the delay occurs when browsing the voice gallery, previewing voices, or generating speech output.
• Approximately how long the loading process takes before the voices become available.

3. Review Speech service limits and throttling

If multiple requests are being issued simultaneously, Speech services may throttle or queue requests.

Some relevant limits include:

Text-to-Speech transactions are subject to service throughput limits.

Custom Voice deployments have concurrency limits that can cause requests to queue when exceeded.

High request volumes can result in increased latency that may appear as prolonged loading within the portal.

If your workload has recently increased or multiple users are accessing the same resource, this could contribute to the behavior.

4. Long Audio or Streaming Scenarios

If you are working with long-form synthesis or streaming scenarios:

Long Audio synthesis is processed asynchronously and may take longer to complete.

Streaming TTS sessions can occasionally experience delays if text is not provided promptly after the connection is established.

Although these scenarios typically affect synthesis rather than voice discovery, they are worth considering depending on how the voices are being used.

Custom Neural Voice overview / eligibility & setup: https://docs.microsoft.com/azure/cognitive-services/speech-service/custom-neural-voice

Get started with Custom Neural Voice: https://docs.microsoft.com/azure/cognitive-services/speech-service/how-to-custom-voice

Prepare training data: https://docs.microsoft.com/azure/cognitive-services/speech-service/how-to-custom-voice-prepare-data

Create and use your voice model: https://docs.microsoft.com/azure/cognitive-services/speech-service/how-to-custom-voice-create-voice

Custom voice deprecation note: (in Custom voice docs) https://docs.microsoft.com/azure/cognitive-services/speech-service/how-to-custom-voice#migrate-to-custom-neural-voice

TTS with Long Audio API (async workflow) : https://docs.microsoft.com/azure/cognitive-services/speech-service/long-audio-api

Azure Speech in Foundry Tools known issues (TTS streaming 503 issue): https://learn.microsoft.com/azure/ai-services/speech-service/known-issues#active-known-issues-text-to-speech-tts

I Hope this helps. Do let me know if you have any further queries.

Thank you!