MioCodec-25Hz-44.1kHz-v2 GGUF

GGUF conversions of Aratako/MioCodec-25Hz-44.1kHz-v2 (MIT license, 133M params) for use with CrispASR.

MioCodec is a lightweight neural audio codec: encode 44.1 kHz audio to 25 Hz discrete tokens (12800 vocab, 341 bps), decode tokens + speaker embedding back to waveform.

Files

File Size Type Notes
miocodec-v2-44k-f16.gguf 259 MB F16 Full precision decode weights
miocodec-v2-44k-q8_0.gguf 155 MB Q8_0 ~2x compression, cos=0.992 vs F32
miocodec-v2-44k-q4_k.gguf 99 MB Q4_K ~4x compression, cos=0.914 vs F32

All GGUFs contain the decoder-only weights (FSQ quantizer + wave_prenet + wave_decoder + wave_post_net + SnakeBeta upsampler + ISTFTHead). The WavLM encoder is separate.

Architecture

tokens(T) -> FSQ(T, 768) -> wave_prenet 6L Transformer(T, 512)
  -> ConvTranspose1d 2x -> interpolate -> ResNet x2
  -> wave_decoder 8L Transformer+AdaLN-Zero(T', 512) -> ResNet x2
  -> SnakeBeta Upsampler 9x -> ISTFTHead(394) -> 44.1kHz waveform
  • FSQ: Finite Scalar Quantization, levels=[8,8,8,5,5], 12800 codebook
  • Transformer: RoPE (normal/adjacent-pair), SwiGLU FFN, windowed attention
  • AdaLN-Zero: Speaker conditioning via 128-dim global embedding
  • SnakeBeta: x + exp(-beta) * sin^2(exp(alpha) * x) periodic activation
  • ISTFT: exp(log_mag) + cos/sin(phase) -> iFFT overlap-add, "same" padding

Parity (F16 vs Python F32 reference)

Stage cos
fsq_decoded 1.000000
wave_prenet_out (6L Transformer) 1.000000
wave_decoder_out (8L AdaLN Transformer) 1.000000
wave_upsampler_out 1.000000
istft_mag_phase 1.000000
output_waveform (175518 samples) 0.999998

Usage with CrispASR

# Diff harness (verify per-stage parity)
crispasr-diff miocodec miocodec-v2-44k-f16.gguf ref.gguf audio.wav

Conversion

python models/convert-miocodec-to-gguf.py \
  --model Aratako/MioCodec-25Hz-44.1kHz-v2 \
  --output miocodec-v2-44k-f16.gguf --dtype f16

# Quantize
crispasr-quantize miocodec-v2-44k-f16.gguf miocodec-v2-44k-q8_0.gguf q8_0
crispasr-quantize miocodec-v2-44k-f16.gguf miocodec-v2-44k-q4_k.gguf q4_k

License

MIT (same as the original model).

Downloads last month
8
GGUF
Model size
0.1B params
Architecture
miocodec
Hardware compatibility
Log In to add your hardware

8-bit

16-bit

Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support

Model tree for cstr/miocodec-v2-44k-GGUF

Quantized
(2)
this model