MAKE IT YOURS
One binding. Your setup.
Pick WoW, GW2, or Generic under Mode. Choose your model and transcription language, change your recording binding, and adjust sending behavior in Advanced settings.
In v0.3.22 and later, open Mode → Edit game profile to customize spoken triggers, channel commands, and recognition keywords for each game.
In v0.3.23 and later, choose the WoW Forever profile that submits chat with controller A. Use Sending → Submit chat with in the profile editor to choose Enter, Controller A, or no submission; regular WoW keeps Enter. Version 0.3.25 sends A through the existing Steam virtual gamepad and retains the controller selected when push-to-talk starts. Use the physical recording binding for automatic A submission; multiple unmatched Steam outputs require resolving the ambiguity and recording again. Version 0.3.26 follows A with B on the same controller to close chat after sending. The A-then-B sequence has been verified in-game; behavior can still depend on your Steam Input/Proton setup.
Version 0.3.23 also adds optional output muting during recording and the CPU-based Parakeet speech engine. Choose the model type and variant on the Model page; Parakeet detects its supported languages automatically. Version 0.3.23 defaults to experimental SDL 3 controller input. Version 0.3.24 restores Legacy as the default following Legion Go compatibility reports; Diagnostics keeps SDL available as an explicit experimental choice.
Version 0.3.27 adds a read-only Legion Go HID input profile, including controller ID 17ef:61eb. Use the default Legacy reader to test recording bindings and controller diagnostics. This is a candidate fix for missing button input; concurrent operation with Steam/InputPlumber still needs hardware validation. The experimental SDL reader does not use this profile.
Version 0.3.24 adds Sending → Casual chat formatting in the profile editor. Turn it off to preserve transcription capitalization and trailing periods. Advanced settings → Sending also offers Exact binding only, which waits 150 ms and prevents larger button combos from starting dictation. In v0.3.25 it appears at the bottom of Sending, below Spoken punctuation.