Re: [Audacity-devel] CD ripping/burning
A free multi-track audio editor and recorder
Brought to you by:
aosiniao
From: James C. <cr...@in...> - 2008-03-20 20:39:04
|
Thomas Holzmann wrote: > James Crook schrieb: >> Thomas Holzmann wrote: > Though I think it would be enough for a SoC project to integrate > denoising features well... It all depends what you mean by 'integrate really well' and what the 's' in feature(s) covers ;-) > So I could make declick and decrackle first and it would be easy to > add denoise later (I think all three could be too much for a project, > or am I wrong?) If all you were doing for the entire project is wrapping three *existing* audio effects taken from Gnome Wave Cleaner source code as LADSPA effects - with default LADSPA GUIs - and then 'tada', that's the improved noise cleaning feature in Audacity, then NO, that would not be enough for a GSoC :-( . Correction - in my opinion that would not be enough. You'd be doing little more than cutting and pasting existing code. You'd be taking a fully worked out and debugged 'audio processing loop' (the hard part that they've put loads of developer hours into) and pasting into a LADSPA template. Allow two weeks for the first one, that's assuming you've not done one before, two days for each extra one thereafter. It's not enough. It is a whole different ballgame if you have to change the audio code as you move it - if it is in ints and needs to move to floats - or if you were taking assembler code from PPC to Pentium.... or if it was inextricably tied up with pThreads. In that case one effect could be a challenging project in itself. I've not looked into the GWC code. If you haven't looked at the effects code yet and gauged what's involved, you need to. For the moment I am assuming that taking the effects is going to be a very straightforward task. You need to put me right if I'm way off here. When assessing the work involved in a project proposal we do need your help to accurately gauge the work involved. The most demanding part of a de-noising project for Audacity looks to me to be in the interface for manual cleaning. I see from the GWC manual that GWC will place up to 200 markers. Our label track could be used as a marker track. Possibly use the labels as is, or possibly modify their appearance for this role? I think there's already a Vamp plug-in that will find clicks and put down labels where they are - you'd need to look into it. You'd then want a mode where you can click on a label and apply an effect... Or select ten and do ten at once... if you choose to make a manual declicker a part of your proposal. You'll probably want to add a denoising toolbar that has the denoising effects on it, rather than only having them as effects in the effects menus. > I think good denoise/declick/deckrackle functions in Audacity would > be a killer feature in Audacity! I know many guys (including me) > working with Audacity who have to switch to another application > because of the poor denoise/declick/deckrackle functions of Audacity. Yes of course. The denoising we have already is popular. Improving it is definitely something we and our users would like to have. > If you really want it I will also think about additional usability > features but atm I have not much ideas about that (maybe a gui for an > assistant "from LP to CD" or so?). I've started a section, "= Audio Cleaning =", on the http://www.audacityteam.org/wiki/index.php?title=Use_Cases wiki page (it's a page for feature requests that are for specific applications). That section will end up with way more in it than can be done by one person in any reasonable time (look at our feature requests list to see how it happens!). From it, you may be able to extract something useful for your written proposal. My opinion - don't try and do a Wizard to assist with cleaning LPs. From what I'm picking up it doesn't play to your strengths. Done well it's a project in itself, even using the existing code unchanged for the cleaning. Denoising is the kind of thing where a tutorial goes a long way towards getting more people using the feature - but we wouldn't require that - we're very happy to get good code that people can use and that has a reference-style page in the manual. > Someone here at the mailinglist said that it would be more important > to get declick, deckrackle working than denoise because it's really > poorly supported. What do you think? Gale has answered innumerable support questions from stuck users, and knows what's he talking about as regards what questions users ask... He said: > The declick and decrackle functions are actually more important for > cleaning up records and if we are looking to be useful, it is > declicking that desperately wants improving, far more than noise > removal (and decrackle that is entirely missing). Take it as gospel. You are of course free to ignore it. I don't myself record from LPs and I don't have clicks and crackle. I do have noise (from disk drives) and when not using USB mics serious mains hum on the built in sound card. For me personally a better de-noiser is more use than a decrackler or a declicker. Providing the mentor is happy with the work you are doing, there is scope for varying the project during the work. To me this looks likely with denoising. What is quite likely to happen on the GUI side is synergy with work that other GSoC students are doing. This is open source collaborative development after all. It could be that someone else implements some label track feature that is useful in work on the denoising interface. The person doing denoising might contribute by trying out the feature and tweaking it for the problem they are trying to solve. It's something I'd like to see happen, so I'm raising the possibility now, to increase the chances that it does. The key will be good communication between people once the projects get underway. --James. |