I'm afraid that our backend library, libsndfile, does not expose functionality for this. You could however write a wrapper around a io.BytesIO object and open this as the ogg file. Your wrapper could then intercept writes.

from python-soundfile.

ok. i tried many python library to do this. basically i wanna solution for
this http://stackoverflow.com/q/25469161/1263451
could you please give me at least a hint to solve this stackoverflow
question ?

On Thu, Aug 28, 2014 at 8:58 PM, Bastian Bechtold [email protected]
wrote:

I'm afraid that our backend library, libsndfile, does not expose
functionality for this. You could however write a wrapper around a
io.BytesIO object and open this as the ogg file. Your wrapper could then
intercept writes.

—
Reply to this email directly or view it on GitHub
https://github.com/bastibe/PySoundFile/issues/71#issuecomment-53736912.

from python-soundfile.

Alright, so here is a simple solution. It tells you how to encode a block of audio in memory.

from io import BytesIO
import pysoundfile as sf
import numpy as np

# generate some audio data
data = np.random.randn(1000)
# create an audio "file"
file = BytesIO()
# write the audio data to the file
sf.write(data, file, 44100, format='ogg')
# get the file content
content = bytes(file.getbuffer())

Note that this creates a whole file for that chunk of data. Depending on the file format, this might carry a lot of overhead.

For example, creating a WAV file for 1000 values will take 2044 bytes (2 bytes per value, plus 44 bytes for the file header). WAV does not compress the data at all. It merely converts it to int16. A FLAC file with the same content will take anything between 100 to 2100 bytes, depending on how well the values can be compressed (e.g. randn values are incompressible, ones are perfectly compressible). In realistic circumstances, don't expect more than about 2x compression. OGG will take 7319 bytes for 1000 values. However, it will only take 24371 bytes to encode 100000 values, so it will amortize it's header size for longer data lengths.

What you really want is some kind of streaming codec though. FLAC and OGG are not meant for this kind of application.

from python-soundfile.

Thank you so much for ur great time and help. i will definitely implement
this.
thanks again

On Fri, Aug 29, 2014 at 1:34 PM, Bastian Bechtold [email protected]
wrote:

Alright, so here is a simple solution. It tells you how to encode a block
of audio in memory.

from io import BytesIOimport pysoundfile as sfimport numpy as np

generate some audio datadata = np.random.randn(1000)# create an audio "file"file = BytesIO()# write the audio data to the filesf.write(data, file, 44100, format='ogg')# get the file contentcontent = bytes(file.getbuffer())

Note that this creates a whole file for that chunk of data. Depending on
the file format, this might carry a lot of overhead.

For example, creating a WAV file for 1000 values will take 2044 bytes (2
bytes per value, plus 44 bytes for the file header). WAV does not compress
the data at all. It merely converts it to int16. A FLAC file with the
same content will take anything between 100 to 2100 bytes, depending on how
well the values can be compressed (e.g. randn values are incompressible,
ones are perfectly compressible). In realistic circumstances, don't
expect more than about 2x compression. OGG will take 7319 bytes for 1000
values. However, it will only take 24371 bytes to encode 100000 values, so
it will amortize it's header size for longer data lengths.

What you really want is some kind of streaming codec though. FLAC and OGG
are not meant for this kind of application.

—
Reply to this email directly or view it on GitHub
https://github.com/bastibe/PySoundFile/issues/71#issuecomment-53848083.

from python-soundfile.