🚀 New Home Assistant Integration: OpenAI GPT-4o Mini TTS 🎙️

Hey everyone,

Since OpenAI released their new GPT-4o Mini TTS model, I’ve been working on a custom integration to bring it into Home Assistant as a speech provider. :microphone:

This allows you to use high-quality, human-like text-to-speech inside Home Assistant’s Voice Assistants, like Assist, with support for custom voices and speech styles.

:star2: Features

:white_check_mark: Uses GPT-4o Mini TTS – OpenAI’s latest, most natural-sounding TTS model
:white_check_mark: 11 voices (alloy, nova, shimmer, etc.)
:white_check_mark: Customisable speech – affect, tone, pronunciation, pauses, and emotion
:white_check_mark: No YAML required – full setup via UI
:white_check_mark: Works inside Home Assistant’s voice assistant

:wrench: How to Install

It’s available as a HACS custom repository:
:link: GitHub Repo

:one: Add it to HACS → Integrations → Custom Repositories
:two: Search for OpenAI GPT-4o Mini TTS and install
:three: Go to Settings → Voice Assistants
:four: Select GPT-4o Mini TTS as the text-to-speech engine
:five: Choose the same voice you initially set it up with and set custom speech instructions


I’d love feedback from the community on how well it works for you and any feature requests you have! :rocket:

Let me know what you think! :blush:

8 Likes

Just read about the new model and thought i google “home assistant GPT-4o-mini-tts”. First thing it found is this 10 minute old thread :grimacing: :clap:

2 Likes

im makikng a few updates to try and speed it up a little :wink: ill update the repo once I finished my testing.

It’s working great so far! Thanks!
Is it possible to make it work with other languages than english?

1 Like

i’d have to look into it, i think openai said at the moment this is mainly based on english

Anyone wanna donate $3000 to my local open ai fund?

Looks like it is working with german if you change
return ["en"]
to
return ["en", "de"]
in your tts.py.

I think it’ll probably work with more languages…

4 Likes

Now, this is great - but what I need is an STT solution (and not NaBu Casa).

What about Whisper?

Whisper is shut down, unless you host it locally - which requires a ton of powerful HW.

Do you mean Willow?

I must do something wrong. I used HACS to get the repo, configure the integration with a valid API Key and got a positive feedback. But the entity is not available or unknown. Unfortuneatly there is no log entry, at the voice menu 4o mini tts is grey and not selectable.

Are there any prerequisits to fulfil? Maybe Tier-3 or Home Assistant Cloud active or sth. like that?

Can I use this integration as TTS for my automation announcements? If yes, how?

1 Like

Yes, sorry - meant Willow.
Desperately seeking an alternative…

Indeed. When the public Willow server shut down I had just ordered a graphics card, so I’ve still got it running. The code on GitHub still works. Expensive though, even on Ebay.

Have you checked in with the Willow group on Discord? They have been exploring alternatives.

A cheap/free alternative is Google Cloud. I use TTS, STT, Gemini (AI).

2 Likes

We shouldn’t have to ask you to elaborate on how you did set up Google as an STT endpoint? TTS with Google is no problem, but I never found a free Google URI for STT.

you have to use the Google Cloud integration. There you can config both, TTS and STT. Sure you need an account and have to generate an API key.

It doesn’t look very “free” to me:
Review pricing for Text-to-Speech | Google Cloud

In the first column you can see the limits for free use.
If you exceed them, sure it will generate costs. I have really lot of announcements every day and never reached the limits for the free tier.
To avoid supprises you can configure a budget limit for example with max 1 $ in the Google Cloud console.
You could also activate the 90 day trial period and just test it.