Using Azure neural TTS services as training data for in-house ASR models

Question

Hi,

given that the quality of neural TTS is very close to human, I wonder whether it is permitted to create audio with it and then use it to improve ASR models? These models would then be used in production in a speech-to-text service.

I tried finding something to this respect in the TOS and I only found a direct reference in the TOS for free-tier accounts, but nothing when we have an azure account and use the pay-as-you-go model.

Thanks in advance for the reply.

Answer 1

In the meantime I managed to get an answer from someone inside Azure that pointed me to the section titled "Limit on customer use of service output" in the Azure cognitive services terms document.

As it is not easy to find and there was no further answer in this thread I thought I would also post it here in case it helps someone else.