Repository navigation
Releases
Releases · rodri45l/opencode-dictate
0.1.2 — hands-free dictation in the terminal
Hands-free dictation inside the terminal, with nothing to install.
Highlights
- Bundled recorder (miniaudio, MIT-0) for macOS arm64/x64 and Linux x64 — no
ffmpeg, no Homebrew, no calibration. Windows and Linux arm64 fall back to
ffmpeg/parecord/arecord/sox. - Streaming capture: 16 kHz PCM is read from the recorder's stdout and the WAV
is assembled in memory. No temp files exist at all, and a killed TUI can't strand
a recorder holding the microphone. - Status indicator: a KITT-style scanner or an amplitude wave, showing grey /
red / amber / violet for listening, speaking, transcribing and muted.f7
switches style,f8mutes. - Answer by voice: questions and permission prompts are answered out loud.
- Voice kill switch: "stop" aborts the agent; "stop conversation mode" leaves
hands-free mode — decided by the LLM, not keyword matching.
Fixes worth noting
- ffmpeg on macOS buffers file output in ~256 KB blocks, which left the VAD blind
for the first ~8 seconds of every capture. Capture now reads the stream. - A fresh opencode instance (no session yet) silently swallowed dictation. Voice now
resolves a session on demand and follows the session you're looking at. - Dropped transcripts that contain no words at all ("."), and clips that are
saturated (a knock) rather than speech. - Only one conversation loop can run per instance.
Requires opencode with TUI plugin support and an OpenAI-compatible STT endpoint
(local or cloud).