AFsample2 is a method that improves the prediction of alternative protein conformations and ensembles by introducing random MSA column masking to reduce the influence of co-evolutionary signals. This approach enhances the structural diversity of models generated by the AF2 neural network. AFsample2 was tested on the OC23 dataset, which includes 23 proteins with open and closed conformations. In 17 out of 23 cases, alternate state models improved without compromising the generation of the preferred state. Consistent results were observed in 16 membrane protein transporters, with improvements in 12 out of 16 targets. TM-score improvements to experimental end states were substantial, sometimes exceeding 50%, elevating mediocre scores from 0.58 to nearly perfect 0.98. AFsample2 also increased the diversity of intermediate conformations by 70% compared to the standard AF2 system, producing highly confident models that could potentially be on-path between the two states. Additionally, a novel strategy was proposed to select end-states in generated model ensembles. These solutions could enhance the generation and identification of alternative protein conformations, providing a more comprehensive understanding of protein function and dynamics. Future work will focus on validating the accuracy of these intermediate conformations and exploring their relevance to functional transitions in proteins. AFsample2 was tested on additional datasets, including a dataset of 16 transporter proteins with inward and outward facing conformations. It was able to generate superior models for both states in most cases. AFsample2 also showed effectiveness in modeling fold switchers, such as the modified S6-ribosomal protein, which alters conformations between two different folds. The method was able to generate models with a reasonable level of confidence for both folds. AFsample2 was also tested on a case study involving newly discovered fold switchers, demonstrating its ability to generate high-quality models of both states. The method was found to be effective in generating diverse protein ensembles and identifying conformational states without the aid of experimental reference structures. The results indicate that AFsample2 is a promising method for improving the prediction of alternative protein conformations and ensembles.AFsample2 is a method that improves the prediction of alternative protein conformations and ensembles by introducing random MSA column masking to reduce the influence of co-evolutionary signals. This approach enhances the structural diversity of models generated by the AF2 neural network. AFsample2 was tested on the OC23 dataset, which includes 23 proteins with open and closed conformations. In 17 out of 23 cases, alternate state models improved without compromising the generation of the preferred state. Consistent results were observed in 16 membrane protein transporters, with improvements in 12 out of 16 targets. TM-score improvements to experimental end states were substantial, sometimes exceeding 50%, elevating mediocre scores from 0.58 to nearly perfect 0.98. AFsample2 also increased the diversity of intermediate conformations by 70% compared to the standard AF2 system, producing highly confident models that could potentially be on-path between the two states. Additionally, a novel strategy was proposed to select end-states in generated model ensembles. These solutions could enhance the generation and identification of alternative protein conformations, providing a more comprehensive understanding of protein function and dynamics. Future work will focus on validating the accuracy of these intermediate conformations and exploring their relevance to functional transitions in proteins. AFsample2 was tested on additional datasets, including a dataset of 16 transporter proteins with inward and outward facing conformations. It was able to generate superior models for both states in most cases. AFsample2 also showed effectiveness in modeling fold switchers, such as the modified S6-ribosomal protein, which alters conformations between two different folds. The method was able to generate models with a reasonable level of confidence for both folds. AFsample2 was also tested on a case study involving newly discovered fold switchers, demonstrating its ability to generate high-quality models of both states. The method was found to be effective in generating diverse protein ensembles and identifying conformational states without the aid of experimental reference structures. The results indicate that AFsample2 is a promising method for improving the prediction of alternative protein conformations and ensembles.