numpy
soxr

[all]
voiceclean[pipecat,rnnoise,silero]

[pipecat]
pipecat-ai>=1.0

[rnnoise]
pyrnnoise>=0.4

[silero]
onnxruntime>=1.16
