![](data:image/svg+xml, %3Csvg xmlns='http://www.w3.org/2000/svg' height='64' class='font-weight-bold' style='font: 600 30.11764705882353px "SegoeUI", Arial' width='64'%3E%3Ccircle fill='hsl(112.00000000000001, 41%, 20%)' cx='32' cy='32' r='32' /%3E%3Ctext x='50%25' y='55%25' dominant-baseline='middle' text-anchor='middle' fill='%23FFF' %3ERI%3C/text%3E%3C/svg%3E)
Foundry Tools
Formerly known as Azure AI Services or Azure Cognitive Services is a unified collection of prebuilt AI capabilities within the Microsoft Foundry platform
![](data:image/svg+xml, %3Csvg xmlns='http://www.w3.org/2000/svg' height='64' class='font-weight-bold' style='font: 600 30.11764705882353px "SegoeUI", Arial' width='64'%3E%3Ccircle fill='hsl(28.799999999999997, 15%, 12%)' cx='32' cy='32' r='32' /%3E%3Ctext x='50%25' y='55%25' dominant-baseline='middle' text-anchor='middle' fill='%23FFF' %3EAV%3C/text%3E%3C/svg%3E)
Thanks for sharing the details. Since your dataset upload and validation completed successfully, and the failure happens only after several hours of training, this usually suggests an infrastructure or backend issue in the region rather than a problem with your data or configuration.
To help narrow this down and move toward a resolution, here are a few practical steps you can take:
Try running the training in a different supported Speech region. Custom Neural Voice training availability and backend capacity can vary by region. If possible, re-run the job in another region such as West US or North Europe. If it succeeds there but continues to fail in East US, that would strongly point to a regional service issue.
It’s also a good idea to quickly review your dataset one more time before re-running the training. Even though validation passes, double-check that the dataset size and duration are within documented limits, the audio is clean and single-speaker with consistent sampling, and the transcripts are well aligned and meet consent requirements. This helps rule out any edge-case data issues that might surface only during long training runs.
Lastly, as a quick sanity check, confirm that your Speech resource endpoint and authentication are working as expected. You can do this by requesting a token from the region endpoint to ensure there are no connectivity or authentication issues on your side.
If the training works in another region but consistently fails in East US, that would strongly indicate a regional service problem.
If the problem persists, please share a few more details like (training job IDs, the approximate time when the job stops, the region where you're running it (East US), and the name of the Speech resource) and we’ll be happy to help you further with these details.
Thankyou!