NeuroDynamicRunning daily
Your own dictation setup
Can a home server turn speech into text as well as a cloud service?
A speech-to-text server on your own hardware, connected privately to LexVoca. How the parts fit together, and how to build your own.
The goalPrivate dictation at homeYour recordings never leave your network.
Where it’s atRunning dailyLast checked 2 October 2026.
What it costsYour hardware and powerThe software is free. A graphics card is optional.
How it works
Follow the arrows. Each part has a guide below.
LexVocaRecords your voice on Windows.
A speech serverTurns the recording into text, on your own hardware.
Your appLexVoca types the words where your cursor is.
Results so far
Measured on everyday hardware. Yours will differ.
Being measuredSeconds per minute of speech, with a graphics card
Being measuredSeconds per minute of speech, without one
0Recordings sent outside the home
What we learned
- Graphics memory runs out. Another model on the same card can leave too little room. Unload one first.
- Check one part at a time. Test the server on its own before connecting LexVoca, or you can’t tell which side is wrong.
- Start without a graphics card. It’s slower, but it proves the whole chain works before you add the faster part.
Build your own
The guides, in order.
Start at the top. Each guide shows how hard it is and how long it takes.