I currently have Ollama set up with Home Assistant Cloud as the TTS option. In this blog, https://www.home-assistant.io/blog/2025/09/11/ai-in-home-assistant/ , it says that streaming has been added, but I can't find the option to enable it. I am using a 7800xt as my gpu. How do I check if it's enabled and if it isn't how would I enable it?
#Is TTS streaming added yet or is it still unavailable?
1 messages · Page 1 of 1 (latest)
Open a chat with Assist on the dashboard page in your browser (not in the app).
Ask the agent to tell a long story.
The components you use in the Assist pipeline are already configured for streaming output.
is it off by default?
How quickly does Ollama generate the full text? The process should take at least 4-5 seconds so that you can see the streaming mechanism at work.
Are the characters in the chat window appearing gradually?
Do you have the latest version of HA?
is it off by default?
This works automatically; if all conditions within the pipeline are met, streaming mode will be selected. HA Cloud TTS and Ollama integration meet these conditions.
I have the latest version and the words are appearing gradually
These settings are good right?
Are you sure you opened your browser and not the app?
The app still hasn't finished developing this feature. But if the sound starts immediately after the last character, it means that streaming is also working. There's no need to worry about it anymore.
I thought the point of streaming is that it plays while the generation isn’t fully finished?
Hey, @somber minnow did you ever figure this out? I’m also running into streaming issues where the voice doesn’t start until the text completely finishes.
Nope, still the same
For streaming to work it has to be supported by both your TTS option and your LLM
@clever grail what are you thoughts on this? https://discord.com/channels/330944238910963714/1470098379811061863
I think everything should be capable of streaming TTS.
I'm having issues too with streaming actually happening in my browser (and voice preview edition). I can see tts_start_streaming: true in the assistant logs after the 60 character limit is reached and I'm seeing text appear in the chat window, but it's still waiting for the entire long message to generate before speaking anything.
I'm using the latest piper app, which supposedly supports streaming
Actually - I'm even seeing in the piper logs that it's starting to synthesize text before the full message is complete, which is what we want. But then the full message still isn't playing until everything is done.
it depends on your hardware, as far as I know only satellites support TTS streaming. Mobile phones and browser windows do not support it
Browser streaming support was the first to arrive, back last summer. The app only received support last month.