Currently, it seems that the Elevenlabs integration waits until the entire speech file has been rendered before it plays. This can cause long delays when waiting for a response, as well as timeouts for very long responses.
The elevenlabs API supports streaming text to speech playback. It would be nice if the Home Assistant ElevenLabs integration used this instead of waiting for the entire response to be ready.
Don’t forget to vote for your own FR. 
This would be fantastic. Their v3 model is fantastic and would love to use it in HASS. But without streaming support it’s really too sluggish 
FYI,
Elevenlabs TTS now has streaming in 2025.11. It is not in the main release notes but mentioned in full change log: Full changelog for Home Assistant 2025.11 - Home Assistant
This uses both TTS streaming in Home Assistant in combination with streaming available from ElevenLabs to ensure playback starts really as soon as possible.
Thanks. Is streaming enabled by default? I still have the feeling that it takes a long time from having all text generated to start of Speech.
Yes it is.
I would check the debug logs then for timings and maybe share here.
Also, you can enable debug logging for eleven labs and in there see when it starts receiving text and then sending it.
Note, the AI also needs to be able to “stream” the text. If the AI integration does not return the response until all text has been received then TTS won’t have anything until then.
Idea is that AI starts streaming response (text) back, which then allows it to be sent to TTS so it can start creating the spoken response and return it.