How to use ffmpeg in flask server to convert between audio formats without writing files to disk?

I have successfully managed to use ffmpeg in python to convert the format of some audio files like this:

command = "ffmpeg -i audio.wav -vn -acodec pcm_s16le output.wav"
subprocess.call(command, shell=True)

However I want to do this in memory and avoid saving the input and output files to disk.

I have found the follwing code to do such a thing (https://stackoverflow.com/questions/11301863/passing-pythons-file-like-object-to-ffmpeg-via-subprocess):

command = ['ffmpeg', '-y', '-i', '-', '-f', 'wav', '-']
process = subprocess.Popen(command, stdin=subprocess.PIPE)
wav, errordata = process.communicate(file)

But I am struggeling to use this in my context.

I am receiving the file on a server as part of a multipart/form-data request.

@server.route("/api/getText", methods=["POST"])
def api():
    if "multipart/form-data" not in request.content_type:
        return Response("invalid content type: {}".format(request.content_type))
    # check file format
    file = request.files['file']
    if file:
        print('**found file', file.filename)

Now I have the file as a FileStorage Object (https://tedboy.github.io/flask/generated/generated/werkzeug.FileStorage.html). This object has a stream, which can be accessed by using the read method. So I thought I might be able to use this as input for ffmpeg like so:

f = file.read()
command = ['ffmpeg', '-y', '-i', '-', '-f', 'wav', '-']
process = subprocess.Popen(command, stdin=subprocess.PIPE)
wav, errordata = process.communicate(f)

However this yields the the following error:

AssertionError: Given audio file must be a filename string or a file-like object

I have also tried another approach which I found online, using io.BytesIO, to which I can't find the source anymore:

memfile = io.BytesIO()  # create file-object
memfile.write(file.read())  # write in file-object
memfile.seek(0)  # move to beginning so it will read from beginning

And then trying this again:

command = ['ffmpeg', '-y', '-i', '-', '-f', 'wav', '-']
process = subprocess.Popen(command, stdin=subprocess.PIPE, stdout=subprocess.PIPE)
wav, errordata = process.communicate(memfile)

This gets me the following error:

TypeError: a bytes-like object is required, not '_io.BytesIO'

Does anyone have an idea of how to do this?

Update

The first error message is actually not a error message thrown by ffmpeg. As v25 pointed out correctly in his answer the first approach also returns a bytes object and is also a valid solution.

The error message got thrown by a library (speech_recognition) when trying to work with the modified file. In the unlikely case of someone coming across the same problem here the solution:

The bytes objected returned by ffmpeg (variable wav) has to be turned into a file-like object as the error message implies. This can easily be done like this:

memfileOutput = io.BytesIO(wav)

As per your comment, this appears to have been fixed by:

wav, errordata = process.communicate(memfile.read())

I'm not 100% sure on why passing f here wouldn't work, because:

import io

print ('file.read()', type(file.read()))

memfile = io.BytesIO() 
memfile.write(file.read())  
memfile.seek(0) 

print ('memfile', type(memfile))

print ('memfile.read()', type(memfile.read()))

gives...

file.read() <class 'bytes'> # (f)
memfile <class '_io.BytesIO'>
memfile.read() <class 'bytes'>

So it appears f and memfile.read() in your case are of the same type.

I'm not actually sure why the former would yield, AssertionError: Given audio file must be a filename string or a file-like object.