diff --git a/include/algorithms/public/NMFMorph.hpp b/include/algorithms/public/NMFMorph.hpp
index ccc23c14b..c55d84315 100644
--- a/include/algorithms/public/NMFMorph.hpp
+++ b/include/algorithms/public/NMFMorph.hpp
@@ -93,6 +93,21 @@ class NMFMorph
     mPos = (mPos + 1) % mH.cols();
   }
 
+  void processFrame(ComplexVectorView v, RealVectorView hFrame, double interpolation) {
+    using namespace Eigen;
+    using namespace _impl;
+    MatrixXd W = MatrixXd::Zero(mW1.rows(), mW1.cols());
+    for (int i = 0; i < W.cols(); i++) {
+      ArrayXd out = ArrayXd::Zero(mW2.rows());
+      mOT[i].interpolate(interpolation, out);
+      W.col(i) = out;
+    }
+
+    VectorXd frame = W * asEigen<Matrix>(hFrame);
+    RealVectorView mag1 = asFluid(frame);
+    mRTPGHI.processFrame(mag1, v, mWindowSize, mFFTSize, mHopSize, 1e-6);
+  }
+
 private:
   MatrixXd                      mW1;
   MatrixXd                      mW2;
diff --git a/include/clients/rt/NMFMorphInClient.hpp b/include/clients/rt/NMFMorphInClient.hpp
new file mode 100644
index 000000000..c3e6abc9b
--- /dev/null
+++ b/include/clients/rt/NMFMorphInClient.hpp
@@ -0,0 +1,133 @@
+/*
+Part of the Fluid Corpus Manipulation Project (http://www.flucoma.org/)
+Copyright 2017-2019 University of Huddersfield.
+Licensed under the BSD-3 License.
+See license.md file in the project root for full license information.
+This project has received funding from the European Research Council (ERC)
+under the European Union’s Horizon 2020 research and innovation programme
+(grant agreement No 725899).
+*/
+#pragma once
+
+#include "../../algorithms/public/NMFMorph.hpp"
+#include "../../algorithms/public/NMF.hpp"
+#include "clients/common/BufferedProcess.hpp"
+#include "clients/common/FluidBaseClient.hpp"
+#include "clients/common/ParameterConstraints.hpp"
+#include "clients/common/ParameterSet.hpp"
+#include "clients/common/ParameterTrackChanges.hpp"
+#include "clients/common/ParameterTypes.hpp"
+
+namespace fluid {
+namespace client {
+namespace nmfmorph {
+
+enum NMFFilterIndex {
+  kSourceBuf,
+  kTargetBuf,
+  kAutoAssign,
+  kInterp,
+  kFFT,
+  kMaxFFTSize
+};
+
+constexpr auto NMFMorphParams = defineParameters(
+    InputBufferParam("source", "Source Bases"),
+    InputBufferParam("target", "Target Bases"),
+    EnumParam("autoassign", "Automatic assign", 1, "No", "Yes"),
+    FloatParam("interp", "Interpolation", 0, Min(0.0), Max(1.0)),
+    FFTParam<kMaxFFTSize>("fftSettings", "FFT Settings", 1024, -1, -1),
+    LongParam<Fixed<true>>("maxFFTSize", "Maxiumm FFT Size", 16384, Min(4),
+                           PowerOfTwo{}));
+
+class NMFMorphInClient : public FluidBaseClient, public AudioOut {
+
+public:
+  using ParamDescType = decltype(NMFMorphParams);
+
+  using ParamSetViewType = ParameterSetView<ParamDescType>;
+  std::reference_wrapper<ParamSetViewType> mParams;
+
+  void setParams(ParamSetViewType& p) { mParams = p; }
+
+  template <size_t N>
+  auto& get() const
+  {
+    return mParams.get().template get<N>();
+  }
+
+  static constexpr auto getParameterDescriptors() { return NMFMorphParams; }
+
+  NMFMorphInClient(ParamSetViewType &p)
+      : mParams{p}, mSTFTProcessor{get<kMaxFFTSize>(), 1, 1} {
+    audioChannelsIn(1);
+    audioChannelsOut(1);
+  }
+
+  index latency() { return get<kFFT>().winSize(); }
+
+  void reset() { mSTFTProcessor.reset(); }
+
+  template <typename T>
+  void process(std::vector<HostVector<T>> &input,
+               std::vector<HostVector<T>> &output, FluidContext &c) {
+    assert(audioChannelsOut() && "No control channels");
+    assert(output.size() >= asUnsigned(audioChannelsOut()) &&
+           "Too few output channels");
+    if (get<kSourceBuf>().get() && get<kTargetBuf>().get()) {
+
+      auto &fftParams = get<kFFT>();
+      auto sourceBuffer = BufferAdaptor::ReadAccess(get<kSourceBuf>().get());
+      auto targetBuffer = BufferAdaptor::ReadAccess(get<kTargetBuf>().get());
+      if (!sourceBuffer.valid() || !targetBuffer.valid()) {
+        return;
+      }       
+
+      index rank = sourceBuffer.numChans();
+      if (targetBuffer.numChans() != rank)
+        return;
+      if (sourceBuffer.numFrames() != fftParams.frameSize()) {
+        return;
+      }
+      if (sourceBuffer.numFrames() != targetBuffer.numFrames()) {
+        return;
+      }
+      if (mTrackValues.changed(rank, fftParams.frameSize(),get<kAutoAssign>())) {
+        tmpSource.resize(rank, fftParams.frameSize());
+        tmpTarget.resize(rank, fftParams.frameSize());
+        tmpAct.resize(rank);
+        tmpActMatrix.resize(rank, 1);
+        tmpMagnitude.resize(1, fftParams.frameSize());
+        for (index i = 0; i < rank; ++i) {
+          tmpSource.row(i) = sourceBuffer.samps(i);
+          tmpTarget.row(i) = targetBuffer.samps(i);
+        }
+        mNMFMorph.init(tmpSource, tmpTarget, tmpActMatrix, fftParams.winSize(),
+                       fftParams.fftSize(), fftParams.hopSize(), get<kAutoAssign>() == 1);
+      }
+      
+      mSTFTProcessor.process(mParams, input, output, c, [&](ComplexMatrixView in, ComplexMatrixView out) {
+        algorithm::STFT::magnitude(in, tmpMagnitude);
+        mNMF.processFrame(tmpMagnitude.row(0), tmpSource, tmpAct);
+        mNMFMorph.processFrame(out.row(0), tmpAct, get<kInterp>());
+      });
+    }
+  }
+
+private:
+  ParameterTrackChanges<index, index, index> mTrackValues;
+  STFTBufferedProcess<ParamSetViewType, kFFT, true> mSTFTProcessor;
+  algorithm::NMFMorph mNMFMorph;
+  algorithm::NMF mNMF;
+  RealMatrix tmpSource;
+  RealMatrix tmpTarget;
+  RealMatrix tmpActMatrix;
+  RealMatrix    tmpMagnitude;
+  RealVector   tmpAct;
+};
+} // namespace nmfmorph
+
+using RTNMFMorphInClient = ClientWrapper<nmfmorph::NMFMorphInClient>;
+
+} // namespace client
+} // namespace fluid