Use a speech recognizer to convert speech into text.
A comma-separated list of results are stored in the variable %VOICE. It's possible there may be several results because of deficiencies in the speech-recognition or e.g. environmental noise. To get e.g. the first result, use the Variable Split action and then access %VOICE1.
If the recognition fails or the specified Timeout is exceeded, %VOICE is cleared (detect this condition with the Is Set operator).
Language Model is a hint to the speech-recognizer for fine-tuning the results.
The default Android speech recognizer requires a network connection and therefore has a high-latency. Other recognizers are available e.g. on Play Store.
Tip: it may be necessary to install the Google Search app in order to access this action on some devices.