Windows desktop-app microphone access for dictation.
A current Vulkan-capable GPU driver is recommended; CPU fallback is included.
Controls
One key, three gestures.
Tap
Read the current text selection through the configured Windows voice.
Double-tap
Cancel pending work and stop active speech or dictation.
Hold
Wait for the key to turn blue and play the ready chime, speak English, then release to transcribe locally and type into the originally focused app.
Troubleshooting
Work from the symptom.
A tap does not read anything
Select text before pressing the key.
Open the action settings and use Test Voice.
Confirm an enabled playback voice is selected.
Confirm the Windows default playback device can produce sound.
If the target app is elevated, launch it normally.
Hold-to-dictate does not start
Keep holding until the key turns blue and the ready chime plays.
In Windows Settings → Privacy & security → Microphone, enable microphone access and Let desktop apps access your microphone.
Confirm the intended microphone is the Windows default recording device.
Dictation is slow
The first hold after Stream Deck starts warms the local model. A current GPU driver enables Vulkan acceleration on compatible hardware; otherwise the plugin selects its bundled CPU fallback automatically. No Vulkan SDK or CUDA toolkit is required.
Text is not inserted after transcription
Keep the original text field focused until the blue processing indicator clears. If focus changes while processing, SpeakDeck refuses to type into the new window. Some browser and Teams fields share one native Windows focus handle and cannot be distinguished from each other.
When reporting a problem, include the Windows version, Stream Deck version, GPU model, what the key displayed, and the relevant log excerpt.
Review before posting
The plugin does not log selected text, microphone audio, or transcripts, but always review a diagnostic file before posting it publicly. Do not include private text, dictated content, credentials, or unrelated machine details.