Views
No views yet
Part of the soniqo.audio speech toolkit — an open, runtime-portable stack for speech AI. This bundle is the LiteRT export, designed to plug into the abstract interfaces inspeech-core(C++ voice-agent orchestration library). Browse all LiteRT bundles in the soniqo LiteRT collection.
| Component | Parameters | Format | Size (INT8) |
|---|---|---|---|
| Encoder (FastConformer) | ~600 M | TFLite | 567.3 MB |
| Decoder + Joint (LSTM + linear) | ~15 M | TFLite | 17.7 MB |
| File | Size | Description |
|---|---|---|
parakeet-encoder.tflite | 567.3 MB | FastConformer encoder, INT8 dynamic weights |
parakeet-decoder-joint.tflite | 17.7 MB | Fused LSTM decoder + joint, INT8 |
vocab.json | 192 KB | 8 192-token SentencePiece vocab |
config.json | 1 KB | Encoder / decoder / joint specs |
audio [1, N] ──► mel fbank (128 bins, 16 kHz) ──► encoder ──► encoded [1, 1024, T']
│
▼
targets (blank-initialized) ──► decoder-joint ──► logits [1, 1, 1, 1030]
│
▼
TDT decode{0, 1, 2, 3, 4} frames. Blank id = 1024, vocab size = 1024, total
logits = 1030 (1024 tokens + 1 blank + 5 durations).Inputs:
audio_signal [1, 128, T] float32 Mel features (log, normalized)
length [1] int64 Valid T (NeMo convention)
Outputs:
encoded [1, 1024, T'] float32 Encoded features
encoded_length [1] int64 Valid T'Inputs:
encoder_out [1, 1, 1024] float32 Current encoder frame
target [1, 1] int64 Last emitted token (blank to start)
h [2, 1, 640] float32 LSTM hidden state
c [2, 1, 640] float32 LSTM cell state
Outputs:
logits [1, 1, 1, 1030] float32 Joint output
h_out [2, 1, 640] float32 Next hidden state
c_out [2, 1, 640] float32 Next cell staten_fft=512, hop_length=160, win_length=400, pre_emphasis=0.97,
log mel with per-utterance normalization. Implement this on the caller
side in native code to match the NeMo reference exactly.STTInterface / TTSInterface / VADInterface / EnhancerInterface; LiteRT implementations plug straight into the interfaces.base_model repository for the full terms.