Get a short email when a new version is released.
Early access · On-device speech recognition
In early access: local speech recognition based on OpenAI Whisper turns the commands you speak in Turkish into drawing operations. Audio is processed on your computer and is not sent to a server; the model of around 190 MB is downloaded once, at your request.
Speech is converted to text on your computer; recordings are not sent to a server.
Listening only starts when you start it: Ctrl+F2 or the microphone button in the Assistant panel; Esc cancels.
The Turkish voice model of around 190 MB is downloaded from the voice settings or on first use; it is not needed for drawing.
Scope note. This feature is in early access. Recognition accuracy depends on the microphone and background noise; you can also type the same commands into the command line.
01Download the voice model
Press the "Download voice model" button in the voice settings, or confirm the prompt that appears the first time you press the microphone button. The Turkish model of around 190 MB is downloaded once from arctecad.com; its size and SHA-256 checksum are verified.
02Start listening
Turn on listening with Ctrl+F2 or the microphone button in the Assistant panel. The microphone only opens when you start it; the status indicator shows the listening, thinking and ready states.
03Say the command
Say supported commands in Turkish. After a short pause, your speech is transcribed on your computer and turned into a drawing operation.
04See the result
The result of the applied operation appears in the Assistant panel. For written requests, the Assistant first shows a plan card; apply it with Apply (Ctrl+Enter) or cancel with Esc.
05Stop
Turn off listening with Ctrl+F2 or the microphone button; Esc cancels the current operation.
Call commands without leaving the mouse and keyboard; you can also type the same commands into the command line.
Speech is recognised on your computer; on in-house projects, no voice data leaves the office.
Once the model is downloaded, speech recognition does not need an internet connection.
Yes, in early access. The speech engine comes with the app; you download the Turkish voice model separately from the voice settings or on first use.
No. Speech is transcribed on your computer; audio is not sent to a server. In a live voice session, audio is processed in memory only; the temporary audio file created for a single-command recording is deleted immediately after processing.
Only to download the voice model once. After that, speech recognition works offline.
Yes. The model is only needed for voice commands; drawing and all other tools work without it.
9 min read
The command names that work on the ARCTE CAD command line, in Turkish and English, with aliases, example syntax and keyboard shortcuts in one table.
Get started quickly, from installation to your first plan, with a Turkish interface, a Turkish command flow and focused tools.
Draw walls, doors and windows as linked architectural objects.
Feature details: Walls and openingsDetect closed boundaries and write the room name and m² area on the plan.
Feature details: Rooms and areasOpen your existing DWG projects and export them as DXF.
Feature details: DWG and DXF exchangeDownload the installer, or choose your plan and get started right away.