Con-CDVAE is a conditional crystal generation method based on the Crystal Diffusion Variational Autoencoder (CDVAE). This paper proposes Con-CDVAE, which extends CDVAE to enable conditional crystal generation. The model introduces new components, a two-step training method, and three generation strategies to improve performance. The effectiveness of Con-CDVAE is demonstrated through extensive testing under various conditions, including single and combined property targets. Ablation studies show the importance of the new components in achieving the model's performance. The physical credibility of the generated crystals is validated through Density Functional Theory (DFT) calculations, confirming Con-CDVAE's potential in material science research. The model uses a diffusion model to generate crystals, with NCSN and DDPM used for crystal generation and latent variable generation, respectively. The model is trained in two steps: first, using CGCNN models to verify generated crystals, and second, using a Prior block to generate latent variables based on given properties. The model includes a Predictor block to improve the latent variable space and a Prior block to generate latent variables with given properties. The model is tested under various conditions, including single and combined property targets, and the results show that the full strategy performs best in generating crystals with specific properties. However, the less strategy is also effective, though not as good as the full strategy. The model is also tested with combination conditions, showing that it can generate crystals with specific formation energy and band gap. The model is validated using DFT calculations, confirming its ability to generate crystals according to target properties. The model is also able to generate crystals outside the training set, showing its potential in material science research. The model is further improved by addressing issues such as symmetry recognition and reducing the MAC and RMSD of generated crystals. The model is also tested with different crystal systems, showing that it can generate crystals with different crystal systems. The model is able to generate crystals with high accuracy, as shown by the DFT validation. The model is also able to generate crystals that are similar to those in the MP database, showing its potential in material science research. The model is further improved by addressing issues such as symmetry recognition and reducing the MAC and RMSD of generated crystals. The model is also tested with different crystal systems, showing that it can generate crystals with different crystal systems. The model is able to generate crystals with high accuracy, as shown by the DFT validation. The model is also able to generate crystals that are similar to those in the MP database, showing its potential in material science research.Con-CDVAE is a conditional crystal generation method based on the Crystal Diffusion Variational Autoencoder (CDVAE). This paper proposes Con-CDVAE, which extends CDVAE to enable conditional crystal generation. The model introduces new components, a two-step training method, and three generation strategies to improve performance. The effectiveness of Con-CDVAE is demonstrated through extensive testing under various conditions, including single and combined property targets. Ablation studies show the importance of the new components in achieving the model's performance. The physical credibility of the generated crystals is validated through Density Functional Theory (DFT) calculations, confirming Con-CDVAE's potential in material science research. The model uses a diffusion model to generate crystals, with NCSN and DDPM used for crystal generation and latent variable generation, respectively. The model is trained in two steps: first, using CGCNN models to verify generated crystals, and second, using a Prior block to generate latent variables based on given properties. The model includes a Predictor block to improve the latent variable space and a Prior block to generate latent variables with given properties. The model is tested under various conditions, including single and combined property targets, and the results show that the full strategy performs best in generating crystals with specific properties. However, the less strategy is also effective, though not as good as the full strategy. The model is also tested with combination conditions, showing that it can generate crystals with specific formation energy and band gap. The model is validated using DFT calculations, confirming its ability to generate crystals according to target properties. The model is also able to generate crystals outside the training set, showing its potential in material science research. The model is further improved by addressing issues such as symmetry recognition and reducing the MAC and RMSD of generated crystals. The model is also tested with different crystal systems, showing that it can generate crystals with different crystal systems. The model is able to generate crystals with high accuracy, as shown by the DFT validation. The model is also able to generate crystals that are similar to those in the MP database, showing its potential in material science research. The model is further improved by addressing issues such as symmetry recognition and reducing the MAC and RMSD of generated crystals. The model is also tested with different crystal systems, showing that it can generate crystals with different crystal systems. The model is able to generate crystals with high accuracy, as shown by the DFT validation. The model is also able to generate crystals that are similar to those in the MP database, showing its potential in material science research.