frame_flutter_stt_host (offline speech-to-text, live captioning)
Connects to Frame, streams audio from the Host (phone) microphone (for now - streaming from Frame mic coming), which is sent through a local (on Host device) Vosk speech-to-text engine (Unfortunately the Flutter package is Android only), and displays the streaming text on the Frame display.
Drop in an alternative Vosk model to perform speech-to-text in a language other than English (vosk-model-small-en-us-0.15 included). Frame only displays languages in a latin character set, so text might need to be tweaked before sending to Frame. The model name appears in main.dart and pubspec.yaml.
As it uses a small (40MB) on-device model, there are limitations in vocabulary. Very long utterances can cause problems (including offscreen text rendering at the moment) so it works best with a short pause between sentences.
Frameshots, Screenshots
https://github.com/user-attachments/assets/ca39afa0-dcdc-43d7-98ce-c002f42d161d
Architecture
