Abstract: Cutting Stock Problems (which are NP-hard), have numerous applications in industry (iron, steel, glass, wood, paper industry) and in studies of computer operations (multi-programmed computer systems, multiprocessor systems). The crucial step in this problem is generating cutting patterns, especially when we have a non-guillotine cutting. This kind of problem also called trim loss problem have several methods: Bottom Left, Best Fit, Difference Process. We propose an algorithm of generating patterns using Difference process method which guarantee improved non-guillotine cutting patterns.