POST large stdin as multipart/form-data with cURL

Telling cURL to directly read a large (several gigabytes) file and POST it as multipart/form-data works:

$ # This works
$ curl localhost -F 'f=@large_file.txt'

However, cURL fails when trying to read the same amount of data from standard input:

$ cat large_file.txt | curl localhost -F 'f=@-'
curl: option -F: is badly used here
curl: try 'curl --help' for more information

(What I want to do, practically, is tar a directory and directly stream that in an HTTP request: tar -cf - large_dir/ | curl localhost -F 'f=@-')

I think this is because cURL stores all of stdin into memory first before sending any data in a request:

-F, --form <name=content>
    ...

    Tell curl to read content from stdin instead of a file by using
    - as filename. This goes for both @ and < constructs. When stdin
    is used, the contents is buffered in memory first by curl to
    determine its size and allow a possible resend. Defining a
    part's data from a named non-regular file (such as a named pipe
    or similar) is unfortunately not subject to buffering and will
    be effectively read at transmission time; since the full size is
    unknown before the transfer starts, such data is sent as chunks
    by HTTP and rejected by IMAP.

Is there any way to have cURL construct a request body in multipart/form-data format as it reads from stdin, and to stream that data to a server, without buffering it in memory or saving it anywhere?

I don't need the Content-Length header to be set.