Here A: Normalization Of Rna Sequencing Counts (By Ercc / Gene Length) is a reply of mine from a few months ago that includes code to take a GTF file and output the GC content and (union gene model) length of each gene in it. You can just remove the GC content stuff. This will give you an easy to use file to trivially convert raw counts to rpkm (take the counts, divide them by 0.001*the length in the output file and then divide by 1 million). It's usually a good idea to use normalized counts if you want the RPKMs to be useful (just import the values into DESeq2 and the counts(something, normalized=T)).

My R code for creating rpkm from HTSeq and GTF file :

First, you should create a list of gene and their length from GTF file by subtracting (column 5) - (column 4) +1, output Tabdelimited will be like :

Gene1   440
Gene2   1200
Gene3   569
  

and another file is HTSeq-count output file which made from SAM/BAM and GTF input files

I have used this code:

file_len<- read.delim("gene_length.txt",header=F,sep="\t")
file_count<- read.delim("sample1.counts",header=F,sep="\t")
colnames(file_len)<- c("GeneName","Len")
colnames(file_count)<- c("GeneName","Count")  
file_count<-file_count[ !grepl("__", file_count$GeneName) ,]
total_count<- sum(file_count$Count)
oneB<-10^9
finallist <- merge(file_len,file_count,by="GeneName")
finallist$RPKM<-0
finallist[,2:4] <- (sapply(finallist[,2:4], as.double))
finallist$RPKM<- (oneB*finallist$Count)/(total_count*finallist$Len)
#finallist<-finallist[finallist$RPKM>1,]
write.table(finallist,file="rpkm.txt",sep="\t", col.names = T, row.names = F)