This paper presents the delay-dependent $$H_\infty$$ H ∞ and generalized H 2 filters design for stochastic neural networks with time-varying delay and noise disturbance. The stochastic neural networks under consideration are subject to time-varying delay in both the state and measurement equations. The aim is to design a stable full-order linear filter assuring asymptotical mean-square stability and a prescribed $$H_\infty$$ H ∞ or generalized H 2 performance indexes for the filtering error systems. Delay-dependent sufficient conditions for the existence of $$H_\infty$$ H ∞ and generalized H 2 filters are both proposed in terms of linear matrix inequalities. Finally, numerical example demonstrates that the proposed approaches are effective.