And another alternative is to combine the UCSC tools bedToGenePred and genePredToGtf.

bedToGenePred in.bed /dev/stdout | genePredToGtf file /dev/stdin out.gtf

You can install these tools with bioconda, or download them here.

Depends on the BED data you want to convert to GTF. If your raw data was originally a GTF file converted with BEDOPS gtf2bed, then the lossless conversion result (BED-formatted) contains all the columns you need to rebuild the original data, by simply printing out columns in a different order and setting the correct coordinate index:

$ gtf2bed < foo.gtf | sort-bed - > foo.bed
$ awk '{print $1"\t"$7"\t"$8"\t"($2+1)"\t"$3"\t"$5"\t"$6"\t"$9"\t"(substr($0, index($0,$10)))}' foo.bed > foo_from_gtf2bed.gtf

Changing the coordinate is really simple, just add one to the start coordinate. But the GTF format also needs attributes such as gene_id and transcript_id fields that are not present in the BED format. Therefore you will need a third source of information.

Alliteratively you could convert bed to gtf using this kscriptlet:

kscript https://git.io/vbJ4B my.bed > my.gtf

Most of the tools do not produce a featured-complete GTF file. Here is one option written in Rust if you require all the fields of a GTF to be completed: bed2gtf