There is also this guide from Seurat, which shows you how to do the pre-processing in R and velocity analysis in python using Jupyter notebooks (also possible using Rstudio).

The key point here is that you convert the Seurat object to an h5seurat object, which is then converted to an h5ad object; this h5ad object is what is then loaded into python to project RNA velocity on Seurat's UMAP embeddings. DON'T try to convert a Seurat object using as.loom() because the function will inevitably fail since the Seurat object does not contain the necessary information to create the loom file (the loom file is generated with information about spliced and unspliced counts, which the Seurat object does not normally contain).

You can use the RunSCVELO function in SCP(https://github.com/zhanghao-njmu/SCP)

library(SCP)
data("pancreas1k")
pancreas1k <- RunSCVELO(srt = pancreas1k, group_by = "SubCellType", liner_reduction = "PCA", nonliner_reduction = "UMAP", return_seurat = TRUE)
ClassDimPlot(pancreas1k, group.by = "SubCellType", reduction = "UMAP", pt.size = NA, velocity = "stochastic")
ClassDimPlot(pancreas1k, group.by = "SubCellType", reduction = "UMAP", pt.size = 5, pt.alpha = 0.2, velocity = "stochastic", velocity_plot_type = "stream")