Solved: Google TTS problem with large numbers

danielattard · 04-18-2023 05:16 PM

I am using Google TTS and running into this error:

Google has returned an error: 400 This request contains sentences that are too long.

with this text which is only 155 bytes:

The Assessment Review Board of Canada has determined that the assessment of 185 Corkstown Road in Ottawa should be reduced from 35447892 to 31537798 in the Large Industrial Class and 66985690 total assessed value.

Why am I receiving this error? Could it have something to do with the large numbers?

kolban

Sadly, I am not familiar with the "make.com" system. Google Cloud provides a rich array of services that can be invoked over the Internet. The text to speech is an example of one such service. Google makes APIs available that can be called to request the service function and, when a request is received, Google satisfies that request. The API call for a text to speech request can be found here. As we look at the API request, we see that it has expected parameters which includes the text to verbalize. In the demo/web site, we pasted in what we think is the text we want verbalized and it comes back correctly. However, all this does is exercise the Google Cloud API ... it doesn't tell us anything about how the "Make.com" module either works or the parameters that are actually being passed when you invoke your "make.com" application. About the best I can suggest is to examine very carefully your configuration of the make.com app (scenario?). Be as sure as possible that what you are sending to Google Cloud to verbalize is the actual text you think you are sending. Look also at other parameters and see if you can map those to what are expected by the API. You might also want to reach out to help to the good folks at make.com. Maybe they have a recipe to add debugging to your application to see what is actually being passed in to the Google Text To Speech call.

View solution in original post

kolban

I visited this page ... https://cloud.google.com/text-to-speech

and copied/pasted your text ... and it seemed to respond correctly.

Can you have a try and see what happens at your end? Can we also look at the parameters of your TTS request?

danielattard

The demo worked absolutely fine with my sample text.:

The problem arises when I try to use the Google Cloud Text to Speech module inside a Make.com scenario. Here is the error from that environment:

kolban

Sadly, I am not familiar with the "make.com" system. Google Cloud provides a rich array of services that can be invoked over the Internet. The text to speech is an example of one such service. Google makes APIs available that can be called to request the service function and, when a request is received, Google satisfies that request. The API call for a text to speech request can be found here. As we look at the API request, we see that it has expected parameters which includes the text to verbalize. In the demo/web site, we pasted in what we think is the text we want verbalized and it comes back correctly. However, all this does is exercise the Google Cloud API ... it doesn't tell us anything about how the "Make.com" module either works or the parameters that are actually being passed when you invoke your "make.com" application. About the best I can suggest is to examine very carefully your configuration of the make.com app (scenario?). Be as sure as possible that what you are sending to Google Cloud to verbalize is the actual text you think you are sending. Look also at other parameters and see if you can map those to what are expected by the API. You might also want to reach out to help to the good folks at make.com. Maybe they have a recipe to add debugging to your application to see what is actually being passed in to the Google Text To Speech call.

danielattard

Thanks Kolban. You were certainly on the right track to mention the different parameters that were being passed to the API from the make.com module.

What I noticed was that the error only occurred when I used one of the Neural2 voices such as the en-US-Neural2-A [MALE] voice.

When I switched to one of the other voices such as the en-US-Studio-O [FEMALE] voice everything worked fine with no error:

I found mention of an ongoing internal bug for this issue. My issue is resolved since I don't need the Neural2 voice to work, but it still seems odd that this bug exists.