In superplastic forming, high elongations can be achieved and this method is used to produce complex parts. One of the drawbacks of this method is non-uniform thickness distribution in final parts. In this research a new method is proposed to control thickness distribution in the process and produce a final part by uniform thickness by adding a part to the die. Pure lead is selected in examining whether the method is applicable due to its considerable strain rate dependency in ambient temperature. At first, tensile tests at various strain rates performed on the material to find required mechanical properties. Then the current gas forming process was simulated. In current process, blank is formed to a die with a semispherical cavity by gas pressure. The required pressure to reach to the optimum strain rate was calculated. The optimum strain rate is the one that results the most elongation in the tensile tests. After designing and manufacturing the die a part was produced and experimental and predicted results were compared to verify the simulation. After verifying simulation in current process optimization of the new proposed process have carried out. Since the proposed new method have several parameters, python scripting in conjunction with neural network and genetic algorithm were used to optimize these parameters. At first, several simulation was done for various input parameters by the use of python scripting. Then the results was used to train and verify the neural network. After that the trained neural network was used as the function of the genetic algorithm. After optimizing the parameters, dies were designed and manufactured. Producing a part by new method showed the improvement of thickness distribution comparing with the current process. Minimum of thickness was also increased about 20 percent that causes improvement of product performance and decreasing of part failure in forming.