Bez popisu
Najít soubor
2014-04-22 11:31:13 +02:00
contrib/docker deploy contrib/ 2014-02-24 14:54:10 +01:00
server Add archive.html (not linked anywhere) 2014-04-15 01:35:14 +02:00
.gitignore Merge client and server in a single bottle app 2013-11-26 12:04:43 +01:00
12factors.md Some doc about 12factors compliance status 2014-04-15 01:20:53 +02:00
LICENSE AGPL license and better readme 2014-04-14 19:06:16 +02:00
README.md Readme is clearer 2014-04-22 11:31:13 +02:00

TechRec

A Python2 web application that assist radio speakers in recording their shows. Meant to be simple to install and to maintain.

It basically takes a directory with the continuous recording and create new files "cutting/pasting" with ffmpeg.

Features

  • little system dependencies: python2 and ffmpeg
  • The interface is extremely simple to use
  • You can have nested recording (ie: to record an interview inside of a whole show)
  • There is no user system: any user opening the website will see the complete status of the applications. There is, also, nothing stored in cookie or similar mechanisms. This means that recording a session does not require a browser to remain open, or any kind of persistence client-side: server-side does it all. It also means that authorization must be done on another layer (for example, your webserver could add a Basic Auth)

How does it work

We suppose that you have a continous recording of your radio broadcasting. What techrec does is taking files from this directory and "cutting/pasting" parts of them. This can boil down to something like

ffmpeg -i concat:2014-20-01-00-00.mp3|2014-20-01-00-01.mp3 -acodec copy -ss 160 -t 1840 foo.mp3

Implementation details

It is based on bottle, to get a minimal framework. Simple APIs are offered through it, and the static site uses them.

Jobs are not dispatched using stuff like celery, but with a thin wrapper over multiprocessing.Pool; this is just to keep the installation as simple as possible.

The encoding part is delegated to ffmpeg, but the code is really modular so changing this is a breeze. To be quicker and avoid the quality issues related to reencoding, the codec "copy" is used: this means that input and output must have the same format.