Choose a transcription model
Messages and context
Voice input transcribes with a speech model you choose in Settings → Voice → Transcription. The default is handy-computer/parakeet-unified-en-0.6b-gguf.
T3 Code docs: voice inputsettings.ts
What it’s for
Controlling which local speech model turns your dictation into text.
T3 Code docs: voice input
When to use it
Use it when setting up dictation or switching to a different model. On-device dictation may need a network connection to download the model first.
T3 Code docs: voice inputcomposer
How to use it
In Settings → Voice → Transcription, download a model and choose Use model if it isn't already selected.
T3 Code docs: voice input
Select the languages you dictate in to compare models that support all of them, then pick one; filtering alone doesn't switch your model.
T3 Code docs: voice input
For a model copy outside Handy's model folder or the Hugging Face cache, on web or desktop paste its file or folder path under the model list and choose Use existing.
T3 Code docs: voice input
What’s changed
- Local transcription models are now recommended to you by language, speed, and accuracy.Updated Nov 20
- Recommends local speech models by selected language, speed, and accuracy.Updated Nov 17
- Local speech-to-text setup reuses compatible voice models already on disk.Updated Nov 11
Related settings
Unofficial notes. Not affiliated with Theo or t3.gg.