TTS - Does the api allow access to the Gemini Flash/Pro Preview TTS models?

[stonecourier]

My boss was hoping to trial using this. I could not see them listed in the supported models.

Thanks for your great library. I have found it very useful.

[jochenkirstaetter]

Hi @stonecourier

I gave a demo of that feature last Saturday at GDG Accra, and no it's not (yet) in the SDK. However, it's among the top features to work on these days.

Would you have any specifics that you are looking for?

Cheers

[jochenkirstaetter]

The models are just strings which I provide the Model constants for, for easier use and auto complete while typing. So, those models can be used directly already.

However, the whole "audio" processing and streaming is not implemented just yet.

Any help, samples, references are appreciated.

Cheers

[jochenkirstaetter]

See release notes of v2.6.0 and v2.6.2 where the latest models have been added to the static Model class as constants.

Not sure what exactly you're looking for.

Cheers

[stonecourier]

Currently we are using your AI library interface to have an AI helper for kids doing maths/english. It helps them with their homework.

We were getting it to speak by passing the returned text to a speech engine, but it has limitations so I was hoping that using the preview models we could make it faster and get past some of the limitations. It currently does English and Spanish. Ideally if we could set it to use TTS and also get the audio stream, that would be ideal.

[jochenkirstaetter]

Thanks @stonecourier for the insight of how you are using the package. Very practical and cool. 😍

As mentioned, the new models can be used directly already using the string value in case that the constant has not been added.

However, the audio streaming part has not been implemented yet. I already had a brief look at the Python SDK to get an idea of the implementation.

Hmm, I'll give this issue to Jules and see how it goes. 🤔

Cheers

[stonecourier]

Thanks. I was experimenting with Google Studio last night to see how the audio worked with single speaker audio. Not sure how practical it will be. As soon as the text starts getting long, the time to generate the audio gets really long. Whether this is just using the website and it is better using a paid for key - no idea.

https://aistudio.google.com/generate-speech?_gl=1*13rvup6*_ga*NjUwODUwODMyLjE3NDkwMjgwNjI.*_ga_P1DBVKWT6V*czE3NDkwMjgwNjIkbzEkZzEkdDE3NDkwMjg0NzIkajI1JGwwJGg3MzY4NDQ5Mzg.