vibeus/sherpa-onnx-int8

The model card for the vibe bot audio AI pipeline, includes all the necessary model and bin file

Here is the code base for vibe bot ai pipeline:

• espeak-ng-data

  • speak data for TTS

• Id1_resnet34_LM_main_part.rknn

  • Speaker ID RKNN part model
  • run on NPU
  • part 1

• Id2_resnet34_LM_post.onnx

  • Speaker ID ONNX part model
  • run on CPU
  • part2

• encoder-epoch-30-avg-4.int8.onnx

  • STT encoder model
  • int8

• encoder-epoch-30-avg-4.onnx

  • STT encoder model
  • float 32

• decoder-epoch-30-avg-4.int8.onnx

  • STT decoder model
  • int8

• decoder-epoch-30-avg-4.onnx

  • STT encoder model
  • float32

• eff_word.rknn

  • KWS rknn model
  • run on NPU

• embedding_864.bin

  • KWS comparision samples
  • store as bin file for efficient

• en_GB-rachel-medium.onnx

  • TTS onnx model
  • rum on CPU
  • GB accent, female

• en_GB-rachel-medium.onnx.json

  • TTS onnx model json file

• en_US-joe-medium.onnx

  • TTS onnx model
  • rum on CPU
  • US accent, male

• en_US-joe-medium.onnx.json

  • TTS onnx model json file

• fb_t.bin

  • KWS feat bin

• joiner-epoch-30-avg-4.int8.onnx

  • STT joiner model
  • int8

• joiner-epoch-30-avg-4.onnx

  • STT joiner model
  • float32

• silero_vad.onnx

  • AI vad model
  • Upload silero_vad.onnx

• tokens.txt

  • STT tokens map text

• voxceleb_CAM++_LM.onnx

  • Speaker ID model

• voxceleb_resnet34_LM.onnx

  • Speaker ID model