cai
2026-07-09 57753e1b1e169369bdc83f3945ce5246c844b316
src/main.rs
@@ -2258,8 +2258,8 @@
            rms_threshold: f64_env("CV_VAD_RMS_THRESHOLD", 0.012),
            peak_threshold: f64_env("CV_VAD_PEAK_THRESHOLD", 0.08),
            start_frames: u32_env("CV_VAD_START_FRAMES", 5).max(1),
            end_silence_ms: u64_env("CV_VAD_END_SILENCE_MS", 700).max(100),
            min_speech_ms: u64_env("CV_VAD_MIN_SPEECH_MS", 300).max(1),
            end_silence_ms: u64_env("CV_VAD_END_SILENCE_MS", 400).max(100),
            min_speech_ms: u64_env("CV_VAD_MIN_SPEECH_MS", 250).max(1),
            max_turn_ms: u64_env("CV_VAD_MAX_TURN_MS", 10_000).max(1_000),
            initial_ignore_ms: u64_env("CV_VAD_INITIAL_IGNORE_MS", 500),
        }
@@ -2372,7 +2372,9 @@
            return None;
        }
        let voiced = rms >= self.config.rms_threshold || peak >= self.config.peak_threshold;
        // Align with cb-sdk's energy-based segmentation: peak is diagnostic only,
        // otherwise isolated spikes can keep a turn open until max_turn_ms.
        let voiced = rms >= self.config.rms_threshold;
        if !self.in_speech {
            self.remember_pre_speech_frame(frame);
        }