Would it not be more efficient to create the decoder in-process, then you could feed it the bitstream via a compute buffer, and the video would be rendered to a texture.