Speakilo / OneVoice On-Device Model Set (Snapdragon 8 Gen 2 / SM8550 Hexagon v73 + Tier 2A)

  • Target NPU: Qualcomm Hexagon HTP v73 (soc_model = 43, htp_arch = 73, QNN SDK 2.51.0)
  • Total files: 388 (1.877 GB)
Step Stage Path Hardware
1 Spoken LID ecapa-lid/* NPU (QNN HTP v73)
2a VI ASR (gipformer INT8) asr/vi/* CPU (Sherpa-ONNX)
2b EN/KO ASR (SenseVoice INT8) asr/en/* CPU (Sherpa-ONNX)
2c KO Streaming ASR asr/ko/* CPU (Sherpa-ONNX)
3 Source Readiness source-readiness/* NPU (QNN HTP v73)
4 Translation (Tier 2A Hy-MT2-1.8B-Q4_0-enviko32k) mt/model.gguf CPU (llama.cpp + KleidiAI +dotprod+i8mm)
5a Korean TTS (piper-ko-kss KSS) tts/piper_ko_kss/* Hybrid CPU (enc_dp_flow) -> NPU QNN HTP v73 (dec_128_2d)
5b Vietnamese TTS (piper-vi-vais1000) tts/piper_vi_vais1000/* Hybrid CPU -> NPU QNN HTP v73 (+ CPU fallback)
5c English TTS (piper-en-lessac) tts/piper_en_lessac/* Hybrid CPU -> NPU QNN HTP v73 (+ CPU fallback)
5d eSpeak-NG data tts/espeak-ng-data/* CPU G2P for EN/VI Piper
Downloads last month
44
GGUF
Model size
2B params
Architecture
hunyuan-dense
Hardware compatibility
Log In to add your hardware

We're not able to determine the quantization variants.

Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support