How to combine LDA (Latent Dirichlet Allocation) as a topic modeling method with Word2vec word embeddings as representation features? Can someone give me an explanation and how to make the code?

I expected results where the combination of LDA and Word embedding was better than the results of topic modeling with conventional LDA alone, I thought of using clustering so that the word2vec method was used for feature representation and then clustering was carried out by looking at the average vector similarity between documents and topics and then using evaluation metrics. silhouette score, but I don’t know yet whether the results will be more effective than just LDA, can you give me suggestions and other best solutions?