Moonshine Tiny DE ONNX
ONNX export of dattazigzag/moonshine-tiny-de optimized for Transformers.js and WebGPU.
- German speech recognition
- Runs fully in-browser
- ~140 MB total (7.7 MB q8 encoder + 116 MB fp32 decoder)
Files:
- onnx/encoder_model_quantized.onnx — q8 quantized encoder (7.7 MB)
- onnx/encoder_model.onnx — fp32 encoder (31 MB)
- onnx/decoder_model_merged.onnx — fp32 decoder (116 MB)
- Downloads last month
- 223