Google cloud transcription Spanish not useful at all

[Sakeador]

The transcription I'm receiving with the Google cloud transcription service in Spanish (MX and CO) languages are not useful at all. More than 40% of the text transcribed is wrong. I have used the https://console.cloud.google.com/speech/transcriptions url with the UI of Speech-to-Text. Not useful at all, tbh. Is there any way of improve the result of this languages?

[kolban]

Looking at this page ... https://cloud.google.com/speech-to-text ... there seems to be a pre-provided sample where you can supply your own audio and see what the transcription return will be.  What would be useful is if you examined the source of your audio and see what quality it contains.  Perhaps find some pre-recorded Spanish recordings and pass those into the sample and see what the resulting quality of output looks like and see if it differs from what you are finding.  Is there something distinct about your audio input ... for example is it highly domain specific and contains rich sets of domain specific jargon or phrases?  Obviously, don't post anything in the least bit sensitive ... but if you can, examples of what you think was said and what Google transcribed it as would be useful.

[Sakeador]

Hello Kolban,

Thank you for your answer. My concern when I published this post was the tool I was using was not really focus on the exact transcription I needed. So I create an instance at https://console.cloud.google.com/speech/transcriptions/list?project=******" and then configured with the same language (Spanish Mexican) used in the audio. After this, followed all the guidelines to the exact detail, but the result was the same, most of the transcription is not useful for me, full of wrong words, so I have to listen myself and write by myself word by word. Now I don't know if I will be charged for a wrong transcription and, moreover, will the service be useful in the future for Mexican audio? In my example, the audio is perfect, is a chat conversation recorded in a studio between 2 mexican guys, so no way the audio would be wrong.

Thank you for your support.