does the HTTP /stream endpoint support ssml parsing

[marble fossil]

AI-generated Answer for suraj_63976

The HTTP /stream endpoint does not currently support SSML parsing. The API processes and combines streamed text in a way that does not recognize SSML tags. However, if you are using WebSockets, SSML parsing can be enabled by using the enable_ssml_parsing query parameter.

For now, if you need SSML support while streaming, it is recommended to use the WebSocket option.

[raven kernel]

I have SSML input but I want to use all the audio in one go , with websocket I'd have to send EOS to get the audio .

[marble fossil]

AI-generated Answer for suraj_63976

The HTTP /stream endpoint does not support SSML parsing. If you need SSML functionality, the WebSocket endpoint allows for SSML parsing by using the enable_ssml_parsing query parameter. However, for scenarios where you want to process SSML input and receive all audio in one go, the WebSocket endpoint might not be ideal since it requires sending an End-of-Stream (EOS) signal to finalize the audio generation.

In this case, you may want to consider using the regular HTTP endpoint instead, as it returns a complete audio file in a single response and is better suited for handling full input text upfront. However, note that SSML parsing is not explicitly mentioned as supported for the regular HTTP endpoint either. If SSML parsing is critical, you may need to use the WebSocket endpoint with the appropriate settings.

[raven kernel]

does the voice_name I provide in the SSML override the voiceId provided in websocket endpoint path ?