Text To Speech selecting different german voice names has no effect

[stetang]

Hi everyone,
since yesterday selecting different german voice names does not result in different spoken voices. 
This seems to regard basic, wavenet and neural2 voice types.

The following voice names all sound like de-DE-Standard-H: -B,-D,-E,-H
The following voice names all sound like de-DE-Standard-G: -A,-C-F,-G 
Voice Types: de-DE-Standard, de-DE-Wavenet and de-DE-Neural2

What I tried:

Demo Page: 
https://cloud.google.com/text-to-speech#demo
Supported voices and languages page:
https://cloud.google.com/text-to-speech/docs/voices

I tried to pick up the sentence used in the supported page:
"Mit Cloud Machine Learning, kann ihre Anwendung Texte, Bilder und mehr interpretieren."
I used this sentence on the demo page and tried to narrow it down by comparing with the supported voices page. 
It's just a guess but it sounds like de-DE-Standard-G and de-DE-Standard-H are supposedly the only voice names used.

I do not get any errors, just not the correct voices.
I checked with using en-US voices, these do result in different spoken voices.
Instead of using /v1beta1/ I also tried to use https://texttospeech.googleapis.com/v1/text:synthesize

Is there something I can do to get the correct voices back?

Demo Page

 Thank you

[MJane · Accepted solution]

Hi @stetang,

Welcome to the Google Cloud Community!

The problem you're experiencing with Google Cloud Text-to-Speech, where different German voice names sound the same, is likely due to an update released in European markets on January 27, 2025.

I understand that this recent update has affected the Text-to-Speech voices, resulting in them having very similar tones, and also making some voices inaccessible which may impact your implementation. With this, I suggest filing a request on the  issue tracker. While I can't provide specific details or timelines at this moment, please keep an eye on the release notes for any updates or new features related to Text to Speech.

You can also refer to this case related to the latest update in Google Cloud Text-to-Speech for additional details.

Cloud Text-to-Speech Voices updating December in European markets

Was this helpful? If so, please accept this answer as “Solution”. If you need additional assistance, reply here within 2 business days and I’ll be happy to help.

[stetang]

hi @MJane, 
thank you for the answer.

If I understand correctly, google has significantly reduced their voice options.
In the Topic link you posted, I found the following link describing the new voice name mappings:
https://services.google.com/fh/files/emails/b_362612430_reminder_1.pdf
This states that all standard, neural2 and wavenet voices do in fact now point to the same voice names: -G or -H.

[JuanVR]

This situation is very dissapointing. In my case the voices were used for npc characters and now I lost part of their identity. Im really surprised about the lack of communication and responsibility regarding the tts service. The examples and documentation are still showcasing the old voices. For how long can i relay in the voices available today? It looks like Im forced to search some other service where they can unsdertand the situations of their clients and not to act against them.