# Per-voice overlap + step lever (merged 75baa3b7)

One voice (cleo, derived baseline overlap 0.42, step 0.20), both
cadence modes, three step settings:

- voice_s-1-*  — steps -1 (overlap 0.22): clipped, pointed delivery
- voice-*      — steps 0 (baseline 0.42)
- voice_s+1-*  — steps +1 (overlap 0.62): legato

Overlap is now a per-voice tunable (VoiceRecord.overlap + .overlapStep);
Speak events carry a steps int so hooks can color a line's delivery.
The old A/B/C sets are anchor points at overlap 0 / 0.5 / 1.