diff --git a/interface/src/gpu/Transform.cpp b/interface/src/gpu/Transform.cpp
index 71b8b0ec92..88922488a4 100644
--- a/interface/src/gpu/Transform.cpp
+++ b/interface/src/gpu/Transform.cpp
@@ -59,20 +59,31 @@ Transform::Mat4& Transform::evalRelativeTransform( Mat4& result, const Vec3& ori
     return result;
 }
 
+void Transform::evalRotationScale(const Mat3& rotationScaleMatrix) {
+    Vec3 scale(glm::length(rotationScaleMatrix[0]), glm::length(rotationScaleMatrix[1]), glm::length(rotationScaleMatrix[2]));
+    if (scale.x < 0.00001f) scale.x = 0.00001f;
+    if (scale.y < 0.00001f) scale.y = 0.00001f;
+    if (scale.z < 0.00001f) scale.z = 0.00001f;
+
+    Mat3 matRotScale(
+        rotationScaleMatrix[0] / scale.x,
+        rotationScaleMatrix[1] / scale.y,
+        rotationScaleMatrix[2] / scale.z);
+    setRotation(glm::quat_cast(matRotScale));
+
+    float determinant = glm::determinant(matRotScale);
+    if (determinant < 0.f) {
+        scale.x = -scale.x;
+    }
+
+    setScale(scale);
+}
+
 void Transform::evalFromRawMatrix(const Mat4& matrix) {
     if ((matrix[0][3] == 0) && (matrix[1][3] == 0) && (matrix[2][3] == 0) && (matrix[3][3] == 1.f)) {
         setTranslation(Vec3(matrix[3]));
 
-        Mat3 matRS = Mat3(matrix);
-
-        Vec3 scale(glm::length(matRS[0]), glm::length(matRS[1]), glm::length(matRS[2]));
-
-        matRS[0] *= 1/scale.x;
-        matRS[1] *= 1/scale.y;
-        matRS[2] *= 1/scale.z;
-
-        setRotation(glm::quat_cast(matRS));
-        setScale(scale);
+        evalRotationScale(Mat3(matrix));
     }
 }
 
@@ -91,18 +102,9 @@ Transform& Transform::mult( Transform& result, const Transform& left, const Tran
     left.updateCache();
 
     result.setTranslation(Vec3(left.getMatrix() * Vec4(right.getTranslation(), 1.f)));
-    
+
     Mat4 mat = left.getMatrix() * right.getMatrix();
-    Mat3 matRS = Mat3(mat);
-
-    Vec3 scale(glm::length(matRS[0]), glm::length(matRS[1]), glm::length(matRS[2]));
-
-    matRS[0] *= 1/scale.x;
-    matRS[1] *= 1/scale.y;
-    matRS[2] *= 1/scale.z;
-
-    result.setRotation(glm::quat_cast(matRS));
-    result.setScale(scale);
+    result.evalRotationScale(Mat3(mat));
 
     return result;
 }
diff --git a/interface/src/gpu/Transform.h b/interface/src/gpu/Transform.h
index 2760241480..bcb353c578 100644
--- a/interface/src/gpu/Transform.h
+++ b/interface/src/gpu/Transform.h
@@ -60,6 +60,7 @@ public:
 
     Transform& evalInverseTranspose(Transform& result);
     void evalFromRawMatrix(const Mat4& matrix);
+    void evalRotationScale(const Mat3& rotationScalematrix);
 
     static Transform& mult( Transform& result, const Transform& left, const Transform& right);
 
@@ -110,6 +111,7 @@ protected:
 };
 
 typedef QSharedPointer< Transform > TransformPointer;
+typedef std::vector< TransformPointer > Transforms;
 
 };
 
diff --git a/interface/src/renderer/Model.cpp b/interface/src/renderer/Model.cpp
index 6627ce3e2b..965f43acb7 100644
--- a/interface/src/renderer/Model.cpp
+++ b/interface/src/renderer/Model.cpp
@@ -554,14 +554,18 @@ bool Model::render(float alpha, RenderMode mode, RenderArgs* args) {
     }
 
     // Let's introduce a gpu::Batch to capture all the calls to the graphics api
-    gpu::Batch batch;
+    _renderBatch.clear();
+    gpu::Batch& batch = _renderBatch;
     GLBATCH(glPushMatrix)();
 
     // Capture the view matrix once for the rendering of this model
-    gpu::TransformPointer viewTransform(new gpu::Transform(Application::getInstance()->getUntranslatedViewMatrix()));
-    viewTransform->postTranslate(Application::getInstance()->getViewMatrixTranslation());
+    if (_transforms.empty()) {
+        _transforms.push_back(gpu::TransformPointer(new gpu::Transform()));
+    }
+    _transforms[0]->evalFromRawMatrix(Application::getInstance()->getUntranslatedViewMatrix());
+    _transforms[0]->postTranslate(Application::getInstance()->getViewMatrixTranslation() + _translation);
 
-    batch.setViewTransform(viewTransform);
+    batch.setViewTransform(_transforms[0]);
 
     GLBATCH(glDisable)(GL_COLOR_MATERIAL);
     
@@ -692,7 +696,6 @@ bool Model::render(float alpha, RenderMode mode, RenderArgs* args) {
     {
         PROFILE_RANGE("render Batch");
         ::gpu::GLBackend::renderBatch(batch);
-        batch.clear();
     }
 
     // restore all the default material settings
@@ -1875,13 +1878,11 @@ int Model::renderMeshes(gpu::Batch& batch, RenderMode mode, bool translucent, fl
                 (const float*)state.clusterMatrices.constData());
 
             gpu::TransformPointer modelTransform(new gpu::Transform());
-            modelTransform->preTranslate(_translation);
             batch.setModelTransform(modelTransform);
         } else {
         //    GLBATCH(glMultMatrixf)((const GLfloat*)&state.clusterMatrices[0]);
 
             gpu::TransformPointer modelTransform(new gpu::Transform(state.clusterMatrices[0]));
-            modelTransform->preTranslate(_translation);
             batch.setModelTransform(modelTransform);
         }
 
diff --git a/interface/src/renderer/Model.h b/interface/src/renderer/Model.h
index 34d6a34c06..ac4f6bde1b 100644
--- a/interface/src/renderer/Model.h
+++ b/interface/src/renderer/Model.h
@@ -36,11 +36,10 @@ class ViewFrustum;
 typedef QSharedPointer<AnimationHandle> AnimationHandlePointer;
 typedef QWeakPointer<AnimationHandle> WeakAnimationHandlePointer;
 
-namespace gpu {
-    class Batch;
-}
 
 #include "gpu/Stream.h"
+#include "gpu/Transform.h"
+#include "gpu/Batch.h"
 
 /// A generic 3D model displaying geometry loaded from a URL.
 class Model : public QObject, public PhysicsEntity {
@@ -283,7 +282,9 @@ private:
     QUrl _url;
 
     gpu::Buffers _blendedVertexBuffers;
-    
+    gpu::Transforms _transforms;
+    gpu::Batch  _renderBatch;
+
     QVector<QVector<QSharedPointer<Texture> > > _dilatedTextures;
     
     QVector<Model*> _attachments;