Silero VAD v5 — MLX MLX compatible weights for Silero VAD v5, converted from the official JIT model. Model Silero VAD v5 is a lightweight (~309K params) voice activity detection model that processes 512 sample chunks (32ms @ 16kHz) with sub millisecond latency. It outputs a speech probability between 0 and 1 for each chunk, with LSTM state carried across chunks for streaming operation. Architecture: STFT → 4×Conv1d+ReLU encoder → LSTM(128) → Conv1d decoder → sigmoid Usage (Swift / MLX) Part of speech swift. Conversion Converts the official Silero VAD v5 JIT model via torch.hub , transposes Conv1d weights for MLX channels last format, sums LSTM biases ( bias ih + bias hh ), and saves as safetensors. Weight Mapping JIT Key MLX Key Shape model.stft.forward basis buffer stft.weight [258, 256, 1] model.encoder.{i}.reparam conv.weight encoder.{i}.weight varies model.encoder.{i}.reparam conv.bias encoder.{i}.bias varies model.decoder.rnn.weight ih lstm.Wx [512, 128] model.decoder.rnn.weight hh lstm.Wh [512, 128] model.decoder.rnn.bias ih + bias hh lstm.bias [512] model.decoder.decoder.2.weight decoder.weight [1, 1, 128] model.decoder.decoder.2.bias decoder.bias [1] License The original Si…
We use cookies for essential functionality and analytics. You can accept or reject analytics cookies.Cookie policy