Talk:More GSoC Ideas from past years

From Audacity Wiki
Revision as of 20:25, 21 March 2008 by James (talk | contribs) (Answer to query from Gale.)
Jump to: navigation, search

Audio Diff

I appreciate this may be irrelevant as the target audience presumably knows, but I am not quite clear how powerful we are saying audio diff is. For example are we saying if we have two takes, one musically perfect but with a sneeze in the audience, another with no audience distraction but a fluffed note at the part of the score where the sneeze was, we can remove the sneeze from the perfect take without removing the music? Or we can do so only if the sneeze occurs at a place where the notes were played properly on both takes (because then the sneeze is the diff)?

  • I've made it a little clearer for someone who isn't already familiar with diff on the notes page. Diff lets you line up the two takes. A mix and match interface could help you select the best fragment for each timeslice, but it won't give you an improvement on the best fragment for that timeslice. If both takes were wrong in the corresponding fragment that's tough.
  • I'm not going to go into it on the diff page, at least not yet, but it IS possible to do a diff of audio with itself. This is a more sophisticated version of what our audio stretching is doing, where it re-uses fragments, and it could allow you to heal audio by using audio nearby, without having to find that audio yourself by hand. Very advanced experimental stuff.


  • I'm thinking of adding a section Difficulty: in addition to Description, Mentors, Skills, Early Spinoff. This would be 'Easy', 'Medium', 'Difficult' and 'Very Difficult', and we could e.g. put 'Medium to Difficult'. I'm unsure about this because most projects can be cut back to be 'Easy', and any project can be extended to be 'Very Difficult', so it may not add much other than a first impression of how hard. James 08:25, 29 February 2008 (PST)
    • Done. James 13:40, 12 March 2008 (PDT)


  • Multi-Channel Audio Support: "Provide multi-channel export" might not be quite clear in context of our current mechanism for exporting to separate channels. Do we mean a GUI to choose 5:1, 7:1...?

Transcription Toolbar

"Some features we have in CVS that are not yet ready for our stable builds include: Transcription ToolBar"

- Just voicing my opinion that it might be a mistake to not continue the current play-at-speed functionality in 1.4 without comparable playback enhancements to replace it. Though you can't change speed during play, it's a feature omission rather than a bug, and I don't think the quite popular functionality that is now offered is currently buggy. - Gale - It's the audio segmentation part of the transcription toolbar that is supremely buggy, and has been disabled by shortening the toolbar to just the speed play. You couldn't run it without long pauses and a crash. We wouldn't remove the play-at-speed feature until we had a replacement. In other investigations I found the current speed-play leaks an entire time-warping data structure every time it is used - which is really very very sloppy, but we get away with it. Probably we need to clarify that it's only part of the transcription toolbar that is being excluded from the stable (and beta) builds. - James.