Zackriya-Solutions/meetily · error · anyhow::Error

Failed to create VAD session: {:?}

Error message

Failed to create VAD session: {:?}

What it means

VadSession::new(config) failed while constructing the silero VAD session used to filter speech before Whisper. The session bundles an ONNX model plus the shown config (pads, min_speech_time, redemption), so construction fails on model/runtime problems or an invalid config combination - most notably a sample rate the model does not support.

Source

Thrown at frontend/src-tauri/src/audio/vad.rs:62

        config.negative_speech_threshold = 0.35;  // Silero default - allows natural pauses

        // CRITICAL FIX: Removed redemption_time capping to support long continuous speech
        // Previous: capped at 400ms, causing VAD to fragment 5-second speech into 40ms segments
        // New: Use full redemption_time from pipeline (2000ms) to bridge natural pauses
        config.redemption_time = Duration::from_millis(redemption_time_ms as u64);
        config.pre_speech_pad = Duration::from_millis(300);   // Pre-speech padding for context
        config.post_speech_pad = Duration::from_millis(400);  // Increased: more context at end

        // CRITICAL FIX: Increased min_speech_time to prevent tiny 40ms fragments
        // Previous: 100ms allowed too-short segments that Whisper rejects
        // New: 250ms ensures segments are substantial enough for Whisper (>100ms requirement)
        config.min_speech_time = Duration::from_millis(250);  // Prevent tiny fragments

        debug!("Creating VAD session with: sample_rate={}Hz, redemption={}ms, min_speech={}ms, input_rate={}Hz",
               VAD_SAMPLE_RATE, redemption_time_ms, 250, input_sample_rate);

        let session = VadSession::new(config)
            .map_err(|e| anyhow!("Failed to create VAD session: {:?}", e))?;

        // VAD uses 30ms chunks at 16kHz (480 samples)
        let vad_chunk_size = (VAD_SAMPLE_RATE as f32 * 0.03) as usize; // 480 samples

        info!("VAD processor created: input={}Hz, vad={}Hz, chunk_size={} samples",
              input_sample_rate, VAD_SAMPLE_RATE, vad_chunk_size);

        Ok(Self {
            session,
            chunk_size: vad_chunk_size,
            sample_rate: input_sample_rate, // Store input rate for resampling ratio in resample_to_16k()
            buffer: Vec::with_capacity(vad_chunk_size * 2),
            speech_segments: VecDeque::new(),
            current_speech: Vec::new(),
            in_speech: false,
            processed_samples: 0,
            speech_start_sample: 0,
            // Initialize state tracking

View on GitHub (pinned to 0281737d87)

Solutions

1. Verify audio is resampled to 16 kHz (VAD_SAMPLE_RATE) before constructing the processor
2. Pin/align the silero VAD and onnxruntime dependency versions (cargo tree -i) and rebuild
3. Reproduce VadSession::new with these exact config values in a minimal binary outside the app
4. Check the packaged app ships the ONNX runtime native libraries

// Smoke-test VAD construction at app startup so failures surface at boot, not mid-recording
fn vad_constructs() -> bool {
    let mut config = VadConfig::new();
    config.sample_rate = VAD_SAMPLE_RATE;
    VadSession::new(config).is_ok()
}

let session = VadSession::new(config).map_err(|e| anyhow!(
    "Failed to create VAD session: {e:?} (sample_rate={}Hz, input={}Hz)",
    VAD_SAMPLE_RATE, input_sample_rate
))?;

When it happens

Trigger: Creating VadProcessor with a config whose sample rate is not one the silero model supports (must be 16 kHz per VAD_SAMPLE_RATE), an ONNX runtime initialization failure (missing native libs in the packaged app), or a dependency version mismatch between the VAD crate and its bundled model.

Common situations: Bundling the app without required onnxruntime libraries; cargo update bumping the VAD/ort crates; passing an input rate that was not resampled to 16k as the pipeline expects.

Related errors

• VAD task panicked: {}
• No speech detected in audio file
• Parakeet transcription failed on segment {}: {}
• VAD processing failed: {}
• No audio samples decoded from file

AI-assisted analysis of Zackriya-Solutions/meetily@0281737d87 (2026-08-16). Data as JSON: /api/errors/9894ea1e0aa5bc53. Report an issue: GitHub.