makepad/libs/voice/tests/fixtures/silero_ref_speech.txt
Admin 0727415daa voice: pure-rust silero v5 vad + learned speech gate in window voice input
Ports the silero v5 16k branch to pure rust (minimal onnx-protobuf weight
extraction, hardcoded graph: stft-conv, 4x conv+relu, lstm cell, sigmoid
head), 512-sample chunks with 64-sample carried context. Validated against
onnxruntime to 2.3e-6 max diff (fixtures committed), ~425us/chunk release.
Model loads from repo-root silero_vad.onnx or MAKEPAD_VAD_MODEL. vad-test
bin for wav files. window_voice_input now gates packets on vad probability
(0.5 enter / 0.35 exit) and falls back to the rms gate when the model file
is missing — the log line says which gate is active.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-07-28 13:47:05 +02:00

94 lines
846 B
Text

0.001670
0.006884
0.008911
0.007857
0.005907
0.005961
0.005976
0.024519
0.027974
0.189549
0.989981
0.988843
0.993698
0.979928
0.965923
0.999283
0.999337
0.999410
0.998439
0.999663
0.999359
0.999934
0.999934
0.999953
0.999978
0.999982
0.999974
0.999986
0.999986
0.999993
0.999980
0.999936
0.999793
0.999920
0.999796
0.997782
0.510518
0.067032
0.020053
0.022056
0.197112
0.982401
0.998537
0.999205
0.999432
0.999953
0.999132
0.999887
0.999988
0.999993
0.999998
0.999996
0.999994
0.999996
0.999998
0.999997
0.999976
0.999993
0.999989
0.999995
0.999998
0.999978
0.999981
0.999998
0.999997
0.999999
0.999996
0.999994
0.999997
0.999999
0.999998
0.999997
0.999997
0.999996
0.999989
0.999937
0.999964
0.999891
0.999835
0.999562
0.991714
0.257646
0.025675
0.010938
0.008862
0.008489
0.007878
0.007458
0.007176
0.006921
0.006759
0.006662
0.006630
0.006659