Audio displayobject #3656
Audio displayobject #3656
Conversation
|
|
||
| # Specifying Audio(url=...) or Audio(filename=...) does not | ||
| # embed the audio data, it only generates an `<audio>` tag with a | ||
| # link to the source. This will not work when offline. |
There was a problem hiding this comment.
filename should embed the data. There is no guarantee that files are actually local (they could be in /tmp, in which case they are not actually accessible with the files url). If someone wants to use files/ url, they should specify that with the url field, rather than the filename field.
|
May I also propose to start considering to have a side library of display-able objects that most of the time are compatible with all IPython version. Of course we could have low level stuff in IPython itself. But with such a library with faster release cycle than IPython , this could allow older version of IPython to use it. |
|
The symantics of the keyword arguments should follow those of the This object should be in |
|
Also, please add an example to the rich display system notebook in |
That was not my main idea. I really think of a repo with lots of display object, even maybe in IPython organisation but out of sync with IPython itself. Of course here audio seem pretty simple (an I have no objection to have it in lib), but where to draw the limit? Those would be nice, are independant of IPython version, and we should consider having one place where those could be added, with why not a monthly release (at least more frequent than IPython itself). |
|
I have moved the component to This means that generated audio will still be WAV but you can include any format you like and your browser supports. I haven't got around to implement |
| Examples | ||
| -------- | ||
| # embedded raw audio data | ||
| Audio(data=2**13*numpy.sin(2*pi*440/44100*numpy.arange(44100)).astype(numpy.int16), |
There was a problem hiding this comment.
I tried this example and it didn't work. Can you try it out and update as needed?
|
So I think the data should be able to be passed as floats as well. It is pretty simple to normalize and convert the int16 internally. |
|
I've fixed the code example in the docstring, it now works for me. However I am still struggling with getting filenames and urls to work as well as with passing floats to the object. Also, the wav-writing stuff assumes to receive a numpy array. Should it be implemented so it accepts any byte string? |
…d of just showing the URL
|
A good example notebook might be http://nbviewer.ipython.org/urls/raw.github.com/Carreau/posts/master/07-the-sound-of-hydrogen.ipynb, that seems to share a lot of the code. You could also use http://nbviewer.ipython.org/url/gist.github.com/filmor/c7ae1a867fc9058ffcd4/raw/91ce69c1400540ed39f68bd92234abfb1dc2ae70/tone-generator.ipynb as a starting point, that incorporates the Python 3 changes and implements a small "synthesizer". |
|
We'll try to review this soon (middle of IPython meeting). |
|
Moving to milestone 2.0 b/c we're too close to release time. This is nice, but it can wait. |
|
1.0 has now been released, back on track. Would be great to have an example notebook. |
|
I looked into that, but there doesn't seem to be a python 3 compatible library to encode audio data. |
|
Like this: http://nbviewer.ipython.org/urls/raw.github.com/ellisonbg/talk-sicm2-2013/master/audio.ipynb On Sat, Aug 10, 2013 at 9:21 AM, Nils Werner notifications@github.comwrote:
Brian E. Granger |
I didn't ment to do it by default, but to also beeing able to do it. wav is quite heavy compared to other format which might be annoying in services like nbviewer. How would be people feeling about merging this and dooing demo-notbook in a subsequent PR ? |
|
No, please do not merge this yet. It is not yet in the state I would like to have it in. |
|
|
Just a quick comment, we'll need to consider cache issues with this element. Otherwise it can be extremely frustrating to use, as browsers cache media very aggressively (on Chrome, even clearing the cache won't play updated audio files whose name hasn't changed, you need to open a new incognito tab or flat out restart the browser). This page has some suggestions on how to achieve no caching. Maybe that needs to be added to the |
|
@nils-werner our usual policy is to close PRs after they sit for more than a month. That doesn't mean we are not interested in the work moving forward - we would love to see this completed. Are you OK with me closing it to keep our PR queue focused on active work? If you finish this up, we would encourage you to reopen the PR to finish review. |
|
Yeah, no problem; I am currently busy doing other stuff anyways. I would love to find others to help me implementing this though as I, while I have experience with web-stuff and audio, have no idea how IPython internally works and how I should implement this. |
|
OK I will close it, but open an issue to track progress with the hopes of getting other people to help out. |
|
Closing this PR until work restarts. I have opened an issue to track progress on the effort: #4241 |
[ not for 1.0 ]
This Pullrequest adds a
class Audio(DisplayObject)to the list of display objects.The object makes use of HTML5's
<audio>tag to render audio data. Right now, the code generates .WAV PCM-data and includes it assrc="data:audio/wav;base64,..."into the HTML.There is currently no way to show the object in Qtconsole or in any other terminal type.
The code that generates the .WAV data is taken from scipy as their implementation is currently only able to write to files and not fill a
ByteIOtype buffer.I have also submitted a PR to scipy that will fix this issue. Once it has been merged (and IPython will be using that version of scipy), lines 730 to 750 can be replaced by a simple call to
scipy.io.wavfile.write(data, buffer, rate).todo :