Speech Studio ACC: S0 resource intermittently enforces F0 15k limit; files marked invalid

Hi A,

Welcome to Microsoft Q&A and Thank you for reaching out.

I understand that you’re experiencing intermittent failures when saving or uploading text files (~17–18k characters) in Speech Studio’s Audio Content Creation (ACC) tool, despite using a Standard (S0) Speech resource.

The issue results in either a 400 error incorrectly citing the F0 tier’s 15,000-character limit or a false “Saved successfully” message, where the file is then marked as invalid (Content format = X). This behavior is inconsistent but reproducible, and it appears to affect files that are well within the 100,000-character limit expected for S0 resources.

I can confirm that this is a known issue affecting certain scenarios in Audio Content Creation. The root cause is an intermittent failure in resource-tier binding, where the system may temporarily associate your session or file operation with an F0-tier context, even though the UI correctly reflects an S0 resource. This causes incorrect quota enforcement and post-save validation failures.

This typically happens due to cached session data, shared resource contexts, or backend service routing misidentifying the selected resource tier. It is more commonly observed in subscriptions with multiple resource tiers or where the ACC interface is accessed across multiple sessions or tabs.

While a permanent fix is being developed, here are a few workarounds that may help reduce or avoid the issue:

Split text files into chunks of ≤14,800 characters.

Use incognito/private browsing mode to reduce session caching conflicts.

Avoid opening multiple ACC projects or sessions in the same browser window.

Refresh or reopen the ACC project before saving to ensure correct resource binding.

Alternatively, consider using the Speech API directly, which allows explicit control of the resource tier.

We’ve reported this to the ACC product engineering team, and a fix is currently under investigation. There’s no action required on your part beyond using the above workarounds.

References

https://learn.microsoft.com/en-us/azure/ai-services/speech-service/speech-services-quotas-and-limits#quotas-and-limits-reference

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

Thank you!

Thanks Sridhar — I really appreciate the detailed explanation and confirmation that this is a known issue.

I’ve been careful to stay well under the 100k character limit, and I’ve verified that my resource is correctly configured as S0 in the Azure Portal. The behavior you described matches exactly what I’m seeing — especially the false “Saved successfully” message followed by Content format = X.

I’ve tried the suggested workarounds (splitting files, incognito mode, refreshing the ACC project), and while they help occasionally, the issue still recurs unpredictably.

Since I’m on an Azure for Students subscription, I can’t open a support ticket to track this formally. Is there any way I can follow progress on the fix or be notified when it’s resolved?

Thanks again for escalating this to the product engineering team.

(Duplicate of my earlier comment — please refer to the reply posted at 3:28 PM for the full response.)

Could you please share the internal tracking ID (ICM/Azure DevOps bug ID) for this issue, and add a “Known issue”/“Under investigation” tag on this thread? I’d like to follow progress and know when the fix ships. If you need tenant/resource details for telemetry, I can provide them privately.

Hi A,

Microsoft’s engineering team is aware of this intermittent issue and is actively working on a fix. Please wait for updates and potential workarounds. As a Student Subscripition you will be having Limited Access.

Thank you!