A method and a device are described for generating a program stream of MPEG coded video and audio signal, wherein several functional tasks are distributed in an optimal manner between hardware on the one hand and software on the other hand. Forming the elementary streams is performed in hardware. Packer and packetiser functions are performed in hardware by pre-packing the elementary streams, filling-in packet header and pack header as much as possible. Parsing of the elementary streams is performed in hardware. Parse information is stored in a meta byte structure in each prepack. Maintaining the P-STD Model is performed in software. Deciding for an audio pack or a video pack is performed in software. Finishing the packet header and pack header is performed in software, on the basis of the parse information present in the meta byte structure in each prepack.