How to send voice recording from HA to a webhook instead of local STT?

[fickle rose]

I'm looking for a way to send recorded voice commands from Home Assistant to a webhook for further processing in n8n. I know there used to be a project called wyoming-webhook that acted as a Wyoming STT server and forwarded audio to an external API/webhook. Unfortunately, the GitHub repo is gone, and I can't find it anymore.

Has anyone here used wyoming-webhook or maybe built an alternative?

My goal is to:
-capture voice commands via microphone in HA (Wyoming Audio),
-send the raw audio to my webhook in n8n,
-forward it to an STT service
-then process the result with an AI agent and return the response to HA as TTS.

[potent storm]

you are describing an assist voice pipeline. why not do this within home assistant?
what are you actually trying to do?

[fickle rose]

Yes, I know this could be done inside Home Assistant, but my use case is a bit different — my server is too weak to run local STT. That’s why I want to send the raw voice recording to an external webhook, where I process it in n8n: send it to STT (ElevenLabs), handle the logic with an AI agent, and send the response back to HA (TTS).

I don’t want to process the voice command locally — I just need a way to capture the audio and send it outside HA. wyoming-webhook used to do that perfectly.

[potent storm]

ElevenLabs has an integration for using it as TTS inside HA
if you use whisper on your other system to do STT you can easily connect that with wyoming.
then just connect the AI to HA using whatever integration you need

so HA goes through the process but does all the outsourcing

https://github.com/rhasspy/wyoming-faster-whisper
can do the STT

voice comes in -> send to whisper -> whisper returns string -> string sent to AI -> AI responds with string -> string send to TTS --> TTS returns audio file --> audio file send back as responce

most use piper instead of ElevenLabs though

TTS runs decent enough even on power power machines. might be better off using external whisper and AI but local piper

[fickle rose]

Exactly, that’s what I was aiming for!
I want to outsource the STT part (to Whisper running externally), while still using Piper locally for TTS — it works fine on my machine.

[potent storm]

yeah, i run whisper on a different system to my home assistant install

got it running with gpu acceleration on my "big" server. although its about all the crappy gpu in there can do at the moment

[fickle rose]

Would you be able to explain step by step what needs to be done to get this running? I’ve been trying to set it up, but I can’t seem to make it work.

Any help or guidance would be amazing 🙏

[potent storm]

https://github.com/rhasspy/wyoming-faster-whisper

run the docker container then connect to it with the wyoming integration

then add it to the pipeline

[fickle rose]

I also got a suggestion to install the Assist Microphone add-on and use it to send recorded audio files directly to a webhook. What do you think about that? Could it be a better or simpler solution compared to setting things up with Wyoming?

[potent storm]

it is not more simple

you run a docker container on your machine you want to do the STT processing then add integration, put in IP and its set up. theres no need to reinvent the wheel

[fickle rose]

I see. I've never set up Docker before, so this is new to me as well.

[potent storm]

maybe i am missing what your getting at and i am sorry if i am, but i feel you are trying to overcomplicate it massively instead of just using the tools that are right in front of you

[fickle rose]

You might be right. But I'm just starting to learn how all of this works and I'm looking for the right solution

[potent storm]

run whisper in docker on your processing system. piper on the home assistant server and whatever ai integration you want/need.

what OS are you running on the system you want to process on?

[fickle rose]

HAOS