Views
No views yet
pip install babelvox1from babelvox import BabelVox
2
3# Models auto-download on first use (~2.5 GB, cached)
4tts = BabelVox(device="NPU", precision="int8",
5 use_cp_kv_cache=True, talker_buckets=[64, 128, 256])
6
7wav, sr = tts.generate("Don't panic.", language="English")
8
9import soundfile as sf
10sf.write("output.wav", wav, sr)1babelvox --device NPU --int8 --cp-kv-cache --talker-buckets "64,128,256" \
2 --text "Hello world" --output hello.wav| Directory | Contents | Size |
|---|---|---|
int8/ | OpenVINO IR models (INT8 quantized) | ~1.7 GB |
weights/ | Numpy embedding tables + projection weights | ~870 MB |
| Component | File | Size | Device |
|---|---|---|---|
| Talker (28L transformer) | talker.xml/.bin | 444 MB | NPU |
| Talker prefill (KV cache) | talker_prefill.xml/.bin | 444 MB | CPU |
| Talker decode (KV cache) | talker_decode.xml/.bin | 444 MB | NPU |
| Code predictor | code_predictor.xml/.bin | 79 MB | CPU |
| CP prefill (KV cache) | cp_prefill.xml/.bin | 79 MB | CPU |
| CP decode (KV cache) | cp_decode.xml/.bin | 79 MB | CPU |
| Speaker encoder | speaker_encoder.xml/.bin | 9 MB | NPU |
| Tokenizer decoder | tokenizer_decoder.xml/.bin | 114 MB | NPU |
| Tokenizer encoder | tokenizer_encoder.xml/.bin | 48 MB | NPU |
| Optimization | RTF | Notes |
|---|---|---|
| FP16 NPU baseline | 3.0x | Full-recompute, 256-token padding |
| + INT8 quantization | 2.1x | These models |
| + CP KV cache | 1.4x | Eliminates redundant CP recomputation |
| + Multi-bucket talker | 1.0x | Real-time speech synthesis |