HOW TO · NOVEMBER 8, 2026

How to talk to an AI assistant without paying speech fees

Run the speech to text and the text to speech on your own computer, as Clarilot does, so no cloud speech service bills you for the time spent listening to you or talking back.

How Clarilot works today, on your home Wi-Fi. Waitlist open.

Where speech fees come from

A voice assistant does two speech jobs on every turn: it turns your voice into text, and it turns the reply text into a voice. When both jobs are sent to a cloud speech service, the service usually charges for the amount of audio it handles. The more you talk, and the longer the answers, the higher the bill.

The alternative: speech on your own computer

Open source engines can do both jobs on an ordinary computer. faster-whisper turns speech into text and Piper turns text into a voice. The local speech guide covers what hardware they need, how to install and test them and the trade offs, with links to each project.

How Clarilot does it

1. You hold the button on the round screen and talk.
2. The round screen sends your voice over your home Wi-Fi to your own computer.
3. Your computer turns the voice into text and passes it to your agent.
4. When the agent is done, your computer turns the final answer into a voice and the round screen plays it.

What still costs money

• The AI model your agent uses. Only the text of each request reaches it, under your own account, exactly as if you had typed it.
• Your own computer, which needs to be on and powerful enough for the speech models you choose.

Trade offs to know

• Speed depends on your computer and the model size you pick. Smaller models are quicker and less accurate.
• Read the license of each Piper voice before you use it in a product; the Piper voices page says some voices are restricted.
• You look after the setup yourself, including updates to the speech engines.

Questions

Why do some voice assistants charge for speech?

Because they send your audio to a cloud speech service, which usually charges for the amount of audio it turns into text or speaks back.

Does Clarilot charge per minute of speech?

No. Speech is turned into text and spoken back on your own computer, so there are no speech fees, however much you use it.

Is anything still sent to the cloud?

The text of each request still goes to the AI model your agent uses, under your own account. Your voice is handled on your own computer.

Which open source tools can do this?

faster-whisper for speech to text and Piper for text to speech. The local speech guide explains both, with links to the projects.

Start from the Clarilot home page, or read the plain guide to speech to text and text to speech on your own computer.

More how to guides: talk to your agents from your pocket · use your agents from a watch · show your agents on the TV.

Join the Clarilot email waitlist for project updates.