mirror of
https://github.com/rowboatlabs/rowboat.git
synced 2026-06-24 20:28:16 +02:00
a12bf4837b
* feat(voice): show audio-reactive waveform instead of live transcript When recording, the chat input now displays only a live waveform that accumulates from the left and grows to full width, with bar heights driven by real mic amplitude. The transcribed words are still captured and submitted, just not shown while recording. New bars animate in and flow smoothly at ~16 updates/sec. * feat(voice): auto-gain waveform bar heights to track voice dynamics Normalize each frame's amplitude against a running peak (instant attack, slow release) at capture time and map it with a near-linear curve, so bar heights accurately reflect how loud/soft the voice is regardless of mic gain — replacing the old fixed-gain sqrt curve that saturated near max. |
||
|---|---|---|
| .. | ||
| cli | ||
| docs | ||
| experimental | ||
| python-sdk | ||
| rowboat | ||
| rowboatx | ||
| x | ||